US Economic Indicators

The general impression reported back to investors after the World Economic Forum Davos 2025 event is that the USD is likely to appreciate against other currencies. A higher USD is good for our AUD-denominated portfolio (we get more AUD per USD).

Exchange Rates - USD/AUD

US Inflation Report: Jan 2025

Over the 12 months to end Jan 2025:

• In Jan 2025, the Consumer Price Index (CPI) for All Urban Consumers (CPI-U) increased 0.7% prior to seasonal adjustment.
• The CPI increased 3.0% to an index level of 317.671 (1982-84=100).
• The CPI for Urban Wage Earners and Clerical Workers (CPI-W) increased 3.0% to an index level of 311.172 (1982-84=100).

In Jan 2025, the index increased 0.7% prior to seasonal adjustment.

Source: US Bureau of Labor Statistics 12 Feb 2025, CNN: 12 Feb 2025

12 Feb 2025: Jerome Powell testified before Congress in relation to the inflation report.

US Advance Monthly Sales for Retail and Food Services

Mar 2025

Advance Estimates of U.S. Retail and Food Services Advance estimates of U.S. retail and food services sales for April 2025, adjusted for seasonal variation and holiday and trading-day differences, but not for price changes, were $724.1 billion, up 0.1% (±0.5%) from the previous month, and up 5.2% (±0.5 points) from April 2024.

Total sales for the Feb 2025 - Apr 2025 period were up 4.8% (±0.5 points) from the same period a year ago. The March 2025 percent change was revised from up 1.5% (±0.5 percent) to up 1.7% (±0.2 percent).

Retail trade sales were down 0.1% (±0.5 points) from Mar 2025, and up 4.7% (±0.5 points) from Mar 2024. Motor vehicle and parts dealers were up 9.4% (±1.8 points) from 2024, while food service and drinking places were up 7.8% (±1.8 points) from April 2024. US Census Bureau: 15 May 2025

Jan 2025: Largest Fall in Nearly 2 Years

Advance estimates of U.S. retail and food services sales for Jan 2025, adjusted for seasonal variation and holiday and trading-day differences, but not for price changes, were $723.9 billion, down 0.9% (±0.5%) from Dec 2024, and up 4.2% (±0.5%) from Jan 2024. Total sales for the Nov 2024 through Jan 2025 period were +4.2% (±0.5%) from the same period a year ago. The Nov 2024 to Dec 2024 change was revised from +0.4% (±0.5%) to +0.7% (±0.3 %). US Census Bureau: 14 Feb 2025

⏏️

Technology trends in AI in 2025

• Gartner - 21 Oct 2024: Top 10 Strategic Technology Trends for 2025

⏏️

Top 20 AI Platforms [31 Dec 2024]

🇺🇸 On 15 Jan 2025, the White House released a Fact Sheet: "Ensuring U.S. Security and Economic Strength in the Age of Artificial Intelligence". A week later (23 Jan 2025) Donald Trump issued an "Executive Order on Removing Barriers to American Leadership in Artificial Intelligence".

ClaudeAI provides the following information about the technical infrastructure used by leading AI companies top 10 AI platforms:

MICROSOFT AND OPENAI

• Computing Infrastructure
  • Azure cloud platform with specialized AI supercomputing
  • Custom-designed AI Optimized Virtual Machines
  • Over 1 million Nvidia GPUs deployed
• Key Technical Components
  • Massive distributed training clusters
  • Azure OpenAI Service integration
  • Custom networking fabric for AI workloads
  • Specialized cooling systems for high-density compute
  • Dedicated fiber networks between datacenters

GOOGLE/ALPHABET

• Computing Infrastructure
  • Cloud TPU v4/v5 custom processors
  • Tensor Processing Units (TPUs)
  • Global network of AI-optimized datacenters
  • Custom-built AI training clusters
• Key Technical Components
  • TensorFlow ecosystem
  • JAX framework
  • Vertex AI platform
  • Custom networking protocol for distributed training
  • Advanced datacenter cooling technologies

NVIDIA

• Computing Infrastructure
  • H100, L40S, and A100 GPU architectures
  • DGX SuperPOD systems
  • Grace Hopper Superchips
  • NVLink interconnect technology
• Key Technical Components
  • CUDA ecosystem
  • TensorRT optimization software
  • NVIDIA AI Enterprise suite
  • Specialized networking fabric
  • Modular datacenter designs

META

• Computing Infrastructure
  • Research SuperCluster (RSC)
  • Custom-designed AI accelerators
  • Large-scale GPU clusters
  • Specialized training infrastructure
• Key Technical Components
  • PyTorch framework
  • Custom networking protocols
  • Distributed training systems
  • Advanced power management
  • Open compute platform designs

AMAZON/AWS

• Computing Infrastructure
  • AWS Trainium chips
  • AWS Inferentia processors
  • Global datacenter network
  • Custom EC2 instances for AI
• Key Technical Components
  • SageMaker platform
  • AWS Neuron SDK
  • Custom networking solutions
  • Proprietary power distribution
  • Advanced cooling systems

ANTHROPIC

• Computing Infrastructure
  • Large-scale GPU clusters
  • Specialized training infrastructure
  • High-performance computing networks
  • Advanced security systems
• Key Technical Components
  • Custom training frameworks
  • Specialized model architectures
  • Advanced safety monitoring systems
  • Proprietary optimization techniques
  • Constitutional AI infrastructure

INTEL

• Computing Infrastructure
  • Gaudi2 AI accelerators
  • Xeon processors with built-in AI
  • Custom ASIC designs
  • Specialized memory systems
• Key Technical Components
  • oneAPI toolkit
  • OpenVINO framework
  • Custom compiler technology
  • Advanced memory hierarchies
  • Specialized interconnects

IBM

• Computing Infrastructure
  • Power10 processors
  • Integrated AI accelerators
  • Quantum-classical hybrid systems
  • Enterprise-grade systems
• Key Technical Components
  • Watsonx platform
  • RedHat OpenShift integration
  • Enterprise security framework
  • Specialized storage systems
  • Advanced memory management

APPLE

• Computing Infrastructure
  • Neural Engine in ⏯️ Apple Silicon
  • Custom ML accelerators
  • Edge AI optimization
  • Specialized mobile processors
• Key Technical Components
  • Core ML framework
  • Custom on-device AI
  • Privacy-preserving computation
  • Specialized mobile architecture
  • Advanced power management

DATABRICKS

• Computing Infrastructure
  • Lakehouse architecture
  • Multi-cloud platform support
  • Distributed processing systems
  • Enterprise-grade security
• Key Technical Components
  • Delta Lake platform
  • MLflow integration
  • Unity Catalog
  • Advanced resource scheduling
  • Multi-tenant architecture

RANKING AI PLATFORMS

OpenLM has listed a leaderboard of AI products/services. OpenLM Chatbot Arena: 9 Apr 2025

The top 20 AI platforms focused on large language models, conversational AI, and related services, based on the most relevant criteria of market capitalization, investment, and influence in the AI space, and [previous rank] as at 15 Feb 2025 (ranked by ClaudeAI) are:

• OpenAI (GPT-4.o)
  • Market value/funding: $29 billion (as of last known valuation)
  • OpenAI, the developer of ChatGPT, is one of the most prominent players in the field, leading advancements in conversational AI, GPT models, and DALL·E (image generation).

• ✅ Google (DeepMind 2.0)
  • Market value/funding: Part of Alphabet (Google's parent company), valued at over USD 1 trillion (entire Alphabet market value)
  • DeepMind is one of the pioneers in AI research and development, working on a variety of models like AlphaGo and the language model PaLM (Pathways Language Model).
  • Alphabet said on 4 Feb 2025 it will spend USD 75 billion on its AI buildout this year, 29% more than Wall Street expected, and investors signaled disappointment at a missed cloud revenue target and began showing impatience over profitability. Reuters: 4 Feb 2025

• Anthropic (Claude 3.0)
  • Market value/funding: $5 billion (as of 2023)
  • Founded by former OpenAI employees, Anthropic is focused on developing large-scale AI models with an emphasis on safety and alignment.

• Microsoft (Azure AI)
  • Market value/funding: Part of Microsoft, valued at over $2 trillion (entire Microsoft market value)
  • Microsoft’s AI services, integrated into its Azure cloud platform, include conversational models, cognitive services, and custom AI solutions.

• ✅ Meta (Meta AI)
  • Market value/funding: $850 billion (Meta overall market value)
  • Meta develops advanced AI technologies, including large models for NLP (e.g., LLaMA), and it integrates these into its social media, VR, and AR platforms.

• Baidu (Chinese language)
  • Market value/funding: $35 billion
  • Baidu, a Chinese tech giant, is heavily involved in AI and has made significant strides with its conversational AI and autonomous vehicle technologies.

• Cohere
  • Market value/funding: $270 million (as of 2023)
  • Cohere focuses on building large-scale language models for enterprise use, and is one of the leading competitors in the NLP space.

• AI21 Labs
  • Market value/funding: $80 million (as of 2023)
  • Based in Tel Aviv, AI21 Labs specializes in AI-driven language models and has built products like Wordtune, an AI writing assistant.

• Mistral
  • Market value/funding: $15 million (as of 2023)
  • Mistral is a startup that is working on open-weight language models with a strong focus on transparency and ease of access for developers.

• Hugging Face
  • Market value/funding: $100 million (as of last funding round)
  • Hugging Face provides a popular hub for AI model sharing and collaboration, including for NLP tasks, and has gained massive traction in the research and developer community.

• IBM Watsonx.ai
  • Market value/funding: Part of IBM, with IBM valued at around $130 billion
  • IBM Watson is one of the oldest and most established players in AI, focusing on enterprise solutions like customer service, data analysis, and industry-specific AI tools.

• X.ai
  • Market value/funding: Acquired by Microsoft for an undisclosed sum in 2023
  • X.ai is known for its AI-powered scheduling assistant, which automates tasks in business communication, integrated into Microsoft services.

• Alibaba DAMO Academy
  • Market value/funding: Part of Alibaba, valued at $500 billion
  • Alibaba’s DAMO Academy is the research arm of Alibaba Group and works on AI models for e-commerce, cloud services, and logistics.
  • ⏭️ Alibaba says that its new artificial intelligence model is superior to DeepSeek. It says the Qwen 2.5 Max edition scored better than Meta Platforms’s Llama and DeepSeek’s V3 model in various tests.

• Replika
  • Market value/funding: Estimated at $50 million (private valuation)
  • Replika focuses on building emotionally intelligent AI companions and has gained popularity for personal conversations.

• Jina AI
  • Market value/funding: $20 million (as of last funding round)
  • Jina specializes in neural search technology and offers solutions for companies looking to integrate conversational AI with their products.

• Rasa
  • Market value/funding: $100 million (as of last funding round)
  • Rasa is an open-source conversational AI company that focuses on building custom, scalable chatbots for enterprises.

• Character.ai
  • Market value/funding: $250 million (as of 2023)
  • Character.ai is a conversational AI platform that focuses on creating customizable characters that can hold engaging conversations with users.

• Vicarious
  • Market value/funding: $150 million (as of last funding round)
  • Vicarious focuses on AI and robotics, using neural network architectures to solve problems related to vision, reasoning, and motor control.

• SoundHound AI
  • Market value/funding: $300 million (market cap as of 2023)
  • SoundHound AI specializes in voice recognition technology and conversational AI, powering voice-activated systems for businesses.

• PaddlePaddle (Chinese language)
  • Market value/funding: Backed by Baidu, so it’s part of Baidu's valuation
  • PaddlePaddle is an open-source deep learning platform that is widely used in China, focusing on natural language processing and AI model development.

Updates:

• OpenAI has rejected a $97.4 billion bid from a consortium led by Elon Musk.Reuters 14 Feb 2025
• Elon Musk says xAI's Grok 3 chatbot to be unveiled on 17 Feb 2025 Reuters 16 Feb 2025
• Amazon has formed a new group focused on agentic AI. Amazon is building its own AI model that incorporates advanced “hybrid reasoning architecture", along the lines of Anthropic’s recently released Claude 3.7 Sonnet. The offering is tentatively scheduled to launch by June 2025 under the Nova brand, a group of generative AI models Amazon unveiled in late 2024. Amazon wants the new model to take a "hybrid reasoning" approach that provides quick answers and more complex extended thinking within a single system. Business Insider 4 Mar 2025

ABOUT OPENAI

The co-founders of OpenAI were Jack Hughes, Greg Brockman, and Elon Musk.

The 9 most senior people in OpenAI were:

• Jack Hughes
• Greg Brockman
• Elon Musk
• Demis Hassabis
• Shane Legg
• Pieter Abbeel
• Patrick Mynyk
• Dario Amodei
• Igor Mordatch

OpenAI's expenditure in the 2021-2022 financial year was USD1.2 billion.

Here's a link to our earliest (not very useful) look at OpenAI and GPT-3 on 4 Nov 2022 (such a short time ago). A few weeks later OpenAI was encouraging people to try Chat_GPT.

AI Revenue and Profit: 2025

Question to OpenAI's ChatGPT

Sam Altman and Elon Musk are debating over a for-profit model for OpenAI. Where does the funding come from for OpenAI's ongoing operations?

ChatGPT response

OpenAI's funding comes from a mix of private investments, partnerships, and revenue generation. Initially, OpenAI was founded as a non-profit with the goal of ensuring artificial intelligence (AI) benefits humanity. However, as the scale of AI research and development grew, the organization transitioned into a "capped-profit" model to attract more funding.

Here are some key sources of funding for OpenAI:

• Initial Investments: OpenAI started with a significant amount of funding from its co-founders, including Elon Musk, Sam Altman, and others. They committed to funding the research, and early investors were crucial for getting the project off the ground.
• Partnership with Microsoft: In 2019, Microsoft invested $1 billion in OpenAI, and later, in 2023, Microsoft invested even more, bringing the total investment to over $13 billion. In return, Microsoft gained exclusive rights to OpenAI’s technologies (such as GPT) for use in their cloud services (Azure) and other products.
• Revenue Generation: OpenAI has monetized some of its offerings, such as the GPT models, through products like ChatGPT Plus, the API (which businesses and developers can use), and enterprise-level services. These provide significant revenue streams to support ongoing operations.
• Research Funding: OpenAI has also received funding through various grants and collaborations with other research institutions and organizations.
• Capped Profit Model: The transition to a capped-profit model allowed OpenAI to raise additional funds from private investors, offering returns, but with a cap on profits to ensure alignment with its mission to benefit humanity. This structure is part of its hybrid model, balancing between non-profit and for-profit elements.

In summary, OpenAI's funding comes primarily from large investments, particularly from Microsoft, a revenue-generating model through product offerings, and private investors who support its work under a unique capped-profit structure.

Next Question

Investments, such as those you have mentioned, need to produce a financial return. How and at what point do OpenAI's investors expect to receive a return on their investments?

ChatGPT response

You're right that investments, especially in a for-profit or capped-profit model like OpenAI's, are made with the expectation of financial returns. OpenAI’s investors, particularly Microsoft and other private backers, expect returns on their investments, but under a unique structure. Here’s how and when they might expect to see those returns:

1. Capped-Profit Model:

Return Cap: The exact cap on returns is not entirely public, but according to OpenAI’s structure, early investors are expected to receive a return of up to 100x on their investment. After reaching that cap, profits from OpenAI's activities would be reinvested into its mission rather than distributed to investors.

2. How Investors Receive Returns:

The primary ways OpenAI's investors would expect to receive returns are through monetizing its AI models and offering services that generate revenue. Some of these ways include:

• Commercial Products: OpenAI generates revenue from offering products like ChatGPT Plus (a subscription service for consumers), the API for developers and businesses, and enterprise services using models like GPT-4. As these products scale, they will generate increasing revenue, which will be shared with investors as per the capped-profit structure.
• Licensing and Partnerships: A major part of the return for investors, particularly Microsoft, comes from exclusive access to OpenAI’s models. Microsoft has integrated OpenAI’s technology into its Azure cloud platform, making it available to a wide range of businesses. As usage of these AI tools grows, Microsoft stands to profit from the sale of Azure services, a portion of which is shared with OpenAI’s investors.
• Equity: If OpenAI were to go public or be acquired by another company, investors could potentially sell their equity stakes in the company. This could be another path for investors to realize a return, although OpenAI has not publicly discussed a potential IPO or acquisition at this time.

3. Timing of Returns:

Short-term: In the short term, investors are likely already seeing some returns through OpenAI’s revenue-generating products, such as the API and business partnerships (e.g., Microsoft’s exclusive access to OpenAI models). As OpenAI continues to scale these services, the company’s revenue will likely grow, generating more returns for its investors.

Medium to Long-term: As OpenAI’s technologies evolve and become more deeply integrated into commercial products, like cloud services or enterprise solutions, its ability to generate revenue will likely increase. This could take several years, with significant returns likely becoming more apparent over a longer time horizon.

Exit Strategy: If OpenAI eventually goes public or is acquired by a larger company, that could offer a more immediate return to investors. However, this exit is not guaranteed and depends on the company's growth trajectory and strategic decisions about its future.

4. Microsoft’s Strategic Interest:

Microsoft’s investment in OpenAI is not just financial. It is also strategic, as Microsoft integrates OpenAI’s AI models into its cloud business. This creates a broader business opportunity for Microsoft to expand its share of the AI and cloud markets. The returns for Microsoft come in the form of long-term growth and revenue generation from Azure, as well as exclusive AI capabilities that enhance their own product offerings. They likely expect financial returns as OpenAI’s AI models become more widely adopted across industries.

In conclusion, OpenAI’s investors expect returns primarily through revenue generated by commercial products and partnerships, with Microsoft benefiting from its exclusive access to OpenAI's models. The timing of returns is tied to OpenAI's growth and the commercialization of its AI technologies. Investors will likely start seeing returns in the medium to long term, depending on the adoption and scalability of OpenAI’s offerings.

Four More AIs Not in the Top 20

Agentic AI

• ✅ Nvidia - 22 Oct 2024: What Is Agentic AI?
• IBM - 11 Oct 2024: Agentic AI: 4 reasons why it’s the next big thing in AI research
• ✅ Nvidia - Dec 2024: ⏯️ Transform Your Business With Agentic AI

Trustworthy AI, XAI

• IBM - 2 Oct 2024: Trustworthy AI: Machine unlearning - Teaching LLMs to forget things
• IBM - Dec 2024: Trustworthy AI: Explainable artificial intelligence (XAI)

Perplexity AI

Perplexity AI is in the news because of its proposal to buy or merge with TikTok from ByteBance. This would be funded by a USD 300 million IPO, with the US Government taking a half stake in the merged business. One key objective is to allow TikTok to continue to be available in the USA and elsewhere.

DeepSeek R1

⏯️ Building DeepSeek - and the man that did it: Liang Wenfeng

In Dec 2024 DeepSeek R1 was reported to be a complex problem-solving model competing with OpenAI’s o1 — which “zoomed to the global top 10 in performance”— yet was built far more rapidly, with fewer, less powerful AI chips, at a much lower cost, according to the Wall Street Journal. Techradar reports that the GPU used by DeepSeek is Nvidia's H800, a cut-down version of the H100. The H800 was launched in Mar 2023.

Barron's reports that DeepSeek topped the Apple App Store in late Jan 2025, and R1 had already cracked the top 10 of the UC Berkeley leader board Chatbot Arena, a ranking of the top-performing AI assistants. Investors’ big worry was about costs. DeepSeek said it had spent just USD 5.6 million training R1, although that headline figure excludes research and development expenses.

DeepSeek said on 27 Jan 2025 it will temporarily limit registrations due to a cyberattack after the company's AI assistant amassed sudden popularity. Reuters 27 Jan 2025

DeepSeek's low cost chip development triggered a selloff of quantum stocks: QBTS, QUBT, RGTI. Semiconductor developer ✅ NVDA was down more than 16%. The impact on NVDA is surprising, given that DeepSeek used NVDA H800 chips.

OpenAI and Microsoft are investigating whether their proprietary technology has been acquired in an unauthorised manner through a technique known as distillation. Guardian

⏏️

Computers Used By Leading AIs

Chips that companies in AI-driven fields typically use include:

• ✅ NVIDIA’s A100 Tensor Core GPUs: A workhorse for AI and deep learning tasks, popular in data centers.
• ✅ Nvidia H100: Part of Nvidia's Hopper architecture, optimized for next-gen AI workloads.
• Tesla V100: Still found in some older setups.
• ✅ Nvidia A40 / A6000: For high-end workstations and AI research.

• ✅ NVIDIA’s A100 Tensor Core GPUs (based on the Ampere architecture) are widely used for training large AI models. These GPUs are designed for deep learning workloads and are known for their high computational power, making them suitable for tasks such as natural language processing (NLP), computer vision, and reinforcement learning.
• ✅ NVIDIA H100 GPUs OpenAI may also use newer GPUs, like the H100 Tensor Core GPUs (based on the Hopper architecture), which are optimized for even more demanding AI tasks, including training large language models and other generative AI systems. These GPUs offer significant improvements in terms of performance and efficiency, especially for high-scale models like GPT.
• TPUs (Tensor Processing Units): In addition to GPUs, OpenAI has also historically used ✅ Google’s TPUs for some workloads. TPUs are specialized processors designed by Google to accelerate machine learning tasks, and while they are often used in Google’s own data centers, they may be part of OpenAI’s infrastructure for certain tasks.
• Custom Hardware and Quantum Computing: OpenAI is also involved in cutting-edge research in quantum computing, and while quantum processors are not yet widely used in mainstream AI models, quantum computing may play a role in future AI research. OpenAI has historically collaborated with organizations like Microsoft, which has its own quantum computing efforts, but quantum processors are still in the early stages for practical AI deployment.

Anthropic

Anthropic, like OpenAI, is likely leveraging cutting-edge hardware for training and deploying its AI models, including the Claude AI family of models. However, specific details about the hardware they use aren't typically disclosed in full. That said, based on industry trends and available information, we can make some educated guesses about the types of processors they likely use for Claude AI:

• ✅ NVIDIA GPUs: Just like OpenAI, Anthropic likely relies heavily on NVIDIA GPUs for training large-scale language models like Claude. The A100 and H100 Tensor Core GPUs are common choices for high-performance AI workloads, due to their high throughput for tensor operations, which are crucial for deep learning. These GPUs excel at handling the parallel processing needed for large neural network models.
• Cloud Providers and Partnerships: Anthropic has partnerships with large cloud providers such as ✅ Google Cloud, ✅ Amazon Web Services (AWS), and possibly ✅ Microsoft Azure. These cloud providers offer scalable GPU resources that are likely used for training and deploying Claude models. Google Cloud, for example, offers both ✅ NVIDIA GPUs and Tensor Core GPUs, which can be used for accelerating deep learning tasks.
• TPUs (Tensor Processing Units): While not as commonly discussed as GPUs, TPUs are another likely candidate for training large-scale AI models. ✅ Google’s TPUs are highly optimized for tensor-based computations and are a natural fit for AI workloads like those of Claude. If Anthropic uses Google Cloud, it's reasonable to assume that TPUs could play a role in their infrastructure.
• Specialized AI Hardware: There’s also a possibility that Anthropic may explore custom AI hardware or future chip architectures designed specifically for scaling AI workloads. This could include hardware developed by specialized firms or in partnership with cloud providers.
• Quantum Computing: Though quantum computing is still in early stages for practical AI applications, like OpenAI, Anthropic is likely to be keeping an eye on quantum computing as well. However, this is more of a long-term research direction, as quantum computing isn't yet at the stage where it can meaningfully accelerate large language model training or deployment.

✅ Google and Lightspeed Venture Partners new investments in Anthropic [NBC 22 Jan 2025]: Google has added USD 1 billion to its USD 2 billion investment in Anthropic [Lightspeed valuation: USD 60 billion]. Google has a 10% stake in Anthropic and a large cloud contract with the company.

Custom Chips Impact on AMD Revenue

Analysts expect AMD's 2024 December-quarter revenue to surge over 22% to USD 7.53 billion, but the company's growth prospects are shadowed by the dominance of ✅ Nvidia in the AI chip market and the rising trend of tech giants like Microsoft, ✅ Amazon.com and ✅ Meta developing custom chips.

Investors see "customer silicon and ✅ Nvidia as the AI chip market going forward," said Ryuta Makino, analyst at AMD investor Gabelli Funds. Customer silicon refers to custom chips designed for specific customers, unlike off-the-shelf processors. Reuters: 3 Feb 2025

Techinvestment Comment: At least a year ago this outcome was identified as a threat to chip makers, whose customers become their competitors. Companies like ✅ Apple have using custom chips to meet their product requirements, rather than building their products around standard chips produced by (for example) ✅ Nvidia and AMD. This caused us to drop AMD from our portfolio, but retain Nvidia.

TSMC Intends to Expand Its Investment in the United States to USD 165 Billion to Power the Future of AI

"TSMC has announced its intention to expand its investment in advanced semiconductor manufacturing in USA by an additional USD 100 billion. Building on the company’s ongoing $65 billion investment in its advanced semiconductor manufacturing operations in Phoenix, Arizona, TSMC’s total investment in USA is expected to reach USD 165 billion. The expansion includes plans for 3 new fabrication plants, 2 advanced packaging facilities and a major R&D team center, solidifying this project as the largest single foreign direct investment in U.S. history." Announcement: 4 Mar 2025 ⏯️ Trump claims credit

Microsoft is developing AI reasoning models to compete with OpenAI

Microsoft's AI division is developing in-house artificial intelligence reasoning models to compete with OpenAI and may sell them to developers. MAI has completed the training of an MAI family of models which perform nearly as well as the leading models from OpenAI (ChatGPT) and Anthropic (Claude) on commonly accepted benchmarks. The team is also training reasoning models, which use chain-of-thought techniques — a reasoning process that generates answers with intermediate reasoning abilities when solving complex problems — that could compete directly with OpenAI. Reuters: 8 Mar 2025

⏏️

Capabilities of quantum computers

A Quantum Computer Buyer's Guide - Quera, 28 Dec 2024:

• Superconducting Qubits: ✅ IBM, ✅ Google
• Trapped Ion: ✅ IonQ, Quantinuum
• Silicon Spin Qubits: ✅ Intel, 🇦🇺 Diraq
• Photonic Qubits: PsiQuantum, Xanadu
• Neutral Atoms: ✅ Quera, Infleqtion

The table below (prepared by OpenAI - Apr 2025) compares current approaches to fault-tolerant quantum computing.

The next table outlines some company products and technology characteristics:

• Quantum technologies are diverse, including photonic, superconducting, trapped-ion, and emerging approaches
• Qubit count ranges from 10+ to 5,000, with different technological approaches
• Fidelity and coherence time vary significantly across platforms
• Data represents the state of quantum computing as of the latest available information. The field is evolving rapidly.

Some issues and terminology:

• Gate fidelity: "Gate fidelity determines how many gates you can run and that determines the size of the algorithms you can run. Like classical computers, logic gates are the basic building blocks of an algorithm. But unlike classical computers, quantum gates aren’t perfect yet, and the errors add up fast." IonQ
• Qubit coherence time: "Coherence time is a measurement of how long a qubit can maintain its complex quantum state — essentially, a qubit’s lifespan. When a qubit is set up in some quantum mechanical state and left alone, how long before that state decays?" IonQ
• Circuit depth: "Circuit depth is the count of time steps from the initialization of the qubits to the final measurement taken. All gates executed in parallel count as one time step, even though the execution times of gates can vary quite a bit. As noted earlier, the T gate is relatively slow, and T-depth is a metric in its own right. But mid-circuit measurements and classical logic, although not implemented in every circuit, are even slower. IonQ
• Quantum Goldilocks Rule: "A large number of qubits isn’t useful if the qubits are of low fidelity (and limited gate depth). A small number of qubits with high fidelity isn’t useful either. A quantum computer that has a sufficient fidelity to allow at least n x n gates (where n is the number of qubits) is just right."IonQ

⏏️

Market competitiveness of currently available quantum computers (December 2024)

AI assessment and ranking of the market competitiveness of the quantum computers listed above (as at 15 Jan 2025):

• Tier 1 - Leading Commercial Systems:
  • ✅ IBM Eagle: Demonstrated strong performance with 127 qubits and good coherence times. IBM's quantum ecosystem and software stack give it a significant practical advantage for real business applications.
    IBM has announced the world’s first quantum processing unit (QPU): the Majorana 1, the world’s first quantum processor powered by topological qubits, harnessing 8 qubits, designed to scale to a million qubits on a single chip IBM 20 Feb 2025. Interesting fact: a majorana particle is a fermion that is its own antiparticle. They were hypothesised by Ettore Majorana in 1937. The term is sometimes used in opposition to Dirac fermion, which describes fermions that are not their own antiparticles.
  • ✅ IonQ Aria/Forte: Ion trap systems showing excellent gate fidelity and connectivity. Strong results in algorithmic benchmarks and growing commercial availability through cloud partnerships.
  • ✅ D-Wave Advantage: While limited to quantum annealing rather than universal quantum computing, it has established clear commercial use cases in optimization problems with thousands of qubits. ⏯️ CEO Alan Baratz interview: 15 Jan 2025. Baratz says that D-Wave is building both gate and annealing quantum computers.
• Tier 2 - Strong Contenders:
  • ✅ Google Sycamore: Achieved quantum supremacy milestone and showed solid performance in research applications, though faced some challenges in scaling up commercial access.
  • ✅ Honeywell H1-1/H1-2: High-fidelity ion trap systems with good coherence times, though with relatively fewer qubits. Now part of Quantinuum.
  • ✅ IBM Hummingbird: While smaller than Eagle, showed good reliability and integration with IBM's quantum software stack.
• Tier 3 - Emerging Players:
  • ✅ Google Willow
  • ✅ IBM Condor
  • ✅ Intel Tunnel Falls
  • ✅ Quantum Computing Inc
  • ✅ Quera Acquila
  • Xanadu Borealis
  • ✅ Microsoft Azure Quantum: Microsoft unveiled its Majorana 1 fermion processor in Feb 2025. This chip is based on a new architecture utilizing topological qubits, which are designed to be more stable and scalable. The Majorana 1 processor represents a significant milestone in Microsoft's pursuit of building a fault-tolerant quantum computer capable of solving complex, real-world problems.
  • PsiQuantum (still in development in Mar 2025)

ChatGPT assessment and ranking of the market competitiveness of the quantum computers listed above (as at 15 Jan 2025):

1. ✅ IBM Eagle (127 qubits): IBM has been a pioneer in quantum computing, offering cloud-based access to its quantum processors through the IBM Quantum Experience platform. The Eagle processor, with 127 qubits, represents a significant advancement in qubit count and performance. IBM's established ecosystem, including the Qiskit software development framework, and collaborations with industry and academia enhance its market competitiveness.
2. ✅ Google Sycamore: Google's Sycamore processor achieved a milestone in quantum supremacy by performing a specific task faster than classical supercomputers. While this demonstrated significant technological capability, Google's quantum computing services are primarily research-focused, with limited direct market offerings compared to IBM.
3. ✅ Honeywell H1-1 and H1-2: Honeywell's quantum computers utilize trapped-ion technology, known for high-fidelity qubits and low error rates. The H1 series is accessible through the cloud via Microsoft's Azure Quantum platform, making it available for commercial use. Honeywell's focus on error correction and scalability contributes to its market competitiveness.
4. ✅ IonQ Aria and IonQ Harmony: IonQ's trapped-ion quantum computers are accessible through cloud platforms like Amazon Braket and Microsoft Azure Quantum. IonQ has been actively increasing qubit counts and improving error rates, positioning its systems as competitive options for various applications.
5. ✅ D-Wave Advantage Series: D-Wave specializes in quantum annealing, which is suited for optimization problems. The Advantage series offers over 5,000 qubits and is available through cloud services like Amazon Braket. While quantum annealing differs from gate-based quantum computing, D-Wave's focus on specific applications provides a competitive niche in the market.
6. ✅ Rigetti Aspen-9: Rigetti's superconducting qubit-based quantum computers are accessible via its Quantum Cloud Services and through platforms like Amazon Braket. The Aspen-9 system offers a 32-qubit processor, and Rigetti's full-stack approach, including software development tools, contributes to its market presence.
7. ✅ Microsoft Azure Quantum: Microsoft provides a cloud platform that offers access to various quantum hardware providers, including IonQ and Honeywell. While Microsoft is developing its own topological qubit technology, its current market competitiveness lies in its role as an integrator, providing a unified platform for quantum computing services.
8. ✅ Intel Tunnel Falls: Intel's Tunnel Falls is a 12-qubit quantum processor based on silicon spin qubits. Intel's approach focuses on leveraging existing semiconductor fabrication infrastructure for scalability. However, as of now, Intel's quantum processors are primarily in the research and development phase, with limited market availability.
9. IQM Garnet: IQM focuses on superconducting qubit technology and aims to provide on-premises quantum computers for specific industrial applications. .
10. ✅ QuEra Aquila: QuEra is developing neutral-atom-based quantum computers, with the Aquila system being a 256-qubit analog quantum simulator launched in Nov 2022. While promising, its technology is in the early stages, and market adoption is limited. However Aquila is available via AWS Braket
11. ✅ QCi Dirac-3: Quantum Computing Inc. (QCi) offers the Dirac-3 system, focusing on quantum-ready software and hardware solutions. However, its market presence and competitiveness are currently limited compared to more established players.
12. Xanadu Borealis: Xanadu's Borealis is a photonic-based quantum processor accessible through the cloud. While it represents an innovative approach, its market competitiveness is currently limited due to the nascent stage of photonic quantum computing technology.

Tier factors [Claude]

1. Demonstrated qubit count and quality (coherence times, gate fidelity)
2. System reliability and uptime
3. Commercial availability and ease of access
4. Software stack maturity
5. Real-world problem-solving capabilities
6. Corporate support and development resources

Quantum Computing Acquisition Strategies

• On premises
• Quantum computing as a service (QCaaS) - ✅ Quera
• Cloud quantum computing, eg ✅ AWS Braket:
  • Gate-based superconducting processors: ✅ IQM, ✅ Rigetti
  • Gate-based ion-trap processors: ✅ IonQ: About gate-based ion-trap processors
  • Neutral atom-based quantum processors: ✅ QuEra Aquila

Case Study: Fidelity Center for Applied Technology (FCAT)

FCAT worked with the ✅ Amazon's Quantum Solutions Lab to create a Proof of Concept that pushes the boundaries of what is possible today with cloud-based quantum computers. The Amazon Braket service allows FCAT developers to develop hardware agnostic code and change quantum computing backends through a single line of code change. This allows FCAT to research the strengths of different quantum backends and build hybrid workflows.

Use Case - Synthetic Security Creation and Option Pricing: FCAT's use case was a custom binary quadratic model (BQM) that evaluates a list of potential stocks to determine which mix of those stocks would yield a synthetic security whose cumulative attributes match those of the specified target index. This BQM was then sent to a quantum annealer to find the optimal composition of this new security. Once the composition was determined, the results were then sent to a gate-based quantum computer which used amplitude estimation to calculate the prices of options on that security. More

Platform: ✅ Amazon Braket is a flexible platform that provides FCAT developers the opportunity to explore the advantages and disadvantages of some currently available quantum computing technologies. It is not clear which quantum computing approach will emerge as the leading solution in the future. FCAT must keep all options open, explore all current solutions simultaneously, and experiment to determine which impact the different approaches have on various use cases.

Team & Tech: The ✅ Amazon Braket service was used for developing the code and accessing the quantum backends. The annealer used was the D-Wave 2000Q system. The gate-based computer was the ✅ Rigetti Aspen-8 on AWS.

⏏️

Examples of Quantum Computers in Actual Use in US Government

This schedule has been extracted from AI searches of quantum computing use cases.

✅ This identifies manufacturers and service providers that are included in our portfolio

✅ Nvidia cuts stake in Arm Holdings, discloses position in China's WeRide

Nvidia reduced its stake in Arm Holdings by about 44% and exited its holdings in Serve Robotics and SoundHound AI in Q4/2024. Reuters 15 Feb 2025

Defense Technology

10 Defense Tech Startups:

• Anduril Industries Inc: Mini fighter jets, autonomous submarines and AI software
• Shield AI: Autonomous drone called V-BAT, along with software called Hivemind
• Applied Intuition: autonomous vehicle driving technology
• Epirus Inc: High-powered anti-drone microwave system
• Hermius: Hypersonic aircraft, backed by Khosla Ventures, Canaan Partners and Sam Altman
• Re:Build Manufacturing: Defence products
• Scale AI: Defense Llama — a version of Meta’s Llama large language model intended for military use
• Skydio: Autonomous drones
• Firefly Aerospace: Lunar landers and Elitra orbital vehicles
• Astranis: The US Space Force selected Astranis to design a fleet of high orbit GPS satellites to thwart the jamming of transmissions, with the goal to launch the first eight by 2028.

Examples of Quantum Computers in the Non-US Government Sector

This schedule has been extracted from AI searches of quantum computing use cases.

✅ This identifies manufacturers and service providers that are included in our portfolio

⏏️

GPU Semiconductors

NVIDIA GB10 Grace Blackwell Superchip, Project DIGITS: Powered by the NVIDIA GB10 Grace Blackwell Superchip, Project DIGITS delivers a petaflop of AI performance in a power-efficient, compact form factor. With the NVIDIA AI software stack preinstalled and 128GB of memory, developers can prototype, fine-tune, and inference large AI models of up to 200B parameters locally, and seamlessly deploy to the data center or cloud.

⏯️ Jensen Huang Announces Project DIGITS

⏏️

News Sources: AI + Quantum + Robotics

Artificial Intelligence

• IEEE Spectrum - Robotics Section: IEEE Spectrum offers in-depth analysis on emerging robotics technologies, key players in the industry, and their commercial applications.
• TechCrunch - Robotics: TechCrunch covers startup ecosystems, funding rounds, and key partnerships in the robotics sector. It focuses on emerging companies and new technologies.
• Elon Musk - 13 Feb 2025: AI chatbot Grok 3 - a challenger to OpenAI's ChatGPT, is in the final stages of development and would be released in a week or two. ⏯️ Reuters
• OpenAI - 14 Feb 2025: OpenAI on Friday rejected a $97.4 billion bid from a consortium led by billionaire Elon Musk for the ChatGPT maker, saying OpenAI is not for sale and that any future bid would be disingenuous. The unsolicited approach is Musk's latest attempt to block the startup he co-founded with OpenAI CEO Sam Altman — but later left — from becoming a for-profit firm, as it looks to secure more capital and stay ahead in the artificial intelligence race. ⏯️ Reuters

• VentureBeat: AI innovations, business strategies, and industry growth. It provides insights into AI startups, funding rounds, and the corporate investments behind AI developments - invaluable for evaluating investment prospects.

Robotics

• TechCrunch - Robotics: TechCrunch covers startup ecosystems, funding rounds, and key partnerships in the robotics sector. It focuses on emerging companies and new technologies.
• Neuralink Update: 4 Mar 2025 Capabilities facilitated by Neuralink brain chips implanted by a surgical robot.

Semiconductors/GPU

• Semiconductor Industry Association:in-depth market analyses, information on key players in the semiconductor space (e.g., TSMC, Intel), and detailed reports on GPU innovations that are central to AI and gaming.

More Updates and News

• Hardware eg: ✅ Intuitive Surgical
• AI for hardware: Vicarious
• Quantum-enhanced AI for robotics: Toyota Research Institute
• Figure AI: Figure AI has had discussions with OpenAI about USD 1.5 billion new funding at a USD 39.5 billion valuation - but this may not happen. Parkway Venture Capital is expected to lead a Series C round. Reuters: 15 Feb 2025

⏏️

Product Announcements

OpenAI Will Add Reasoning AI Model o3 Mini 15 Dec 2024

The o1 models are capable of reasoning through complex tasks and can solve more challenging problems than previous models in science, coding and math, the company had said in a blog post.

The new o3 and o3 Mini models would be more powerful than the previously launched o1 models, the company had said previously.

OpenAI said it was introducing a beta feature called Tasks to ChatGPT, signaling its foray into the virtual assistant space, competing with ✅ Apple's Siri and ✅ Amazon's Alexa. OpenAI: 15 Dec 2024

D-Wave

22 Jan 2025: ✅ D-Wave Announces New Leap Quantum LaunchPad™ Program to Fast-track Deployment of Quantum Computing Applications

... and as of January 2025 there is DeepSeek-R1.

⏏️

Supercomputers

The 9 most powerful supercomputers in Jan 2025 [Keumars Afifi-Sabet, LiveScience]

1. El Capitan:
   • 🇺🇸 Location: Lawrence Livermore National Laboratory — California, USA
   • Performance: 1,742 petaFLOPS (1.742 exaFLOPS)
   • Components: AMD 4th-Gen EPYC 24-core CPUs in AMD Instinct MI300A APUs
   • First online: November 2024
2. Frontier:
   • 🇺🇸 Location: Oak Ridge National Laboratory — Tennessee, USA
   • Performance: 1,353 petaFLOPS (1.4 exaFLOPS)
   • Components: AMD EPYC 64-core CPUs and AMD Instinct MI250X GPUs
   • First online: August 2022
3. Aurora:
   • 🇺🇸 Location: Argonne National Laboratory — Illinois, USA
   • Performance: 1,012 petaFLOPS (1.0 exaFLOPS)
   • Components: Intel Xeon Max Series CPUs and Intel Data Center Max Series GPUs
   • First online: June 2023
4. Eagle:
   • 🇺🇸 Location: Microsoft Azure — The Cloud / Distributed
   • Performance: 561 petaFLOPS (0.56 exaFLOPS)
   • Components: Intel Xeon Platinum 8480C 48C CPUs and Nvidia H100 GPUs
   • First online: August 2023
5. HPC6:
   • 🇺🇸 Location: Microsoft Azure — The Cloud / Distributed
   • Performance: 561 petaFLOPS (0.56 exaFLOPS)
   • Components: Intel Xeon Platinum 8480C 48C CPUs and Nvidia H100 GPUs
   • First online: August 2023
6. Fugaku:
   • 🇯🇵 Location: Riken Center for Computational Science — Kobe, Japan
   • Performance: 442 petaFLOPS (0.44 exaFLOPS)
   • Components: A64FX CPUs
   • First online: June 2020
7. ALPS:
   • 🇨🇭 Location: Swiss National Supercomputing Centre — Lugano, Switzerland
   • Performance: 270 petaFLOPS
   • Components: Nvidia GH200 Grace Hopper processors
   • First online: September 2024
8. LUMI:
   • 🇫🇮 Location: CSC Data Center — Kajaani, Finland
   • Performance: 380 petaFLOPS (0.38 exaFLOPS)
   • Components: AMD 3rd-Gen EPYC 64-core CPUs and AMD Instinct MI250X GPUs
   • First online: June 2021
9. Leonardo:
   • 🇮🇹 Location: CINECA data center — Bologna, Italy
   • Performance: 239 petaFLOPS (0.23 exaFLOPS)
   • Components: Intel Xeon Platinum 8358 32-core CPUs and Nvidia A100 GPUs
   • First online: November 2022

News

Forschungszentrum Jülich announced on 13 Feb 2025 that it is permanently incorporating a D-Wave annealing quantum computer after 3 years of successful hosting and usage. It will remain a key part of the quantum computing infrastructure at Forschungszentrum Jülich and is expected to become increasingly important. The System will be connected to JUPITER supercomputer, Europe’s first and only exascale HPC, to facilitate breakthroughs in artificial intelligence and quantum optimization applications.

Political Issues

2 Feb 2025: UK orders Apple to let it spy on user encrypted data. Donald Trump denies UK order to force Apple to provide it with access to encryted user data. [George Orwell: 1984]

Donald Trump Floats 25% Tariff on Chips

Donald Trump said on 18 Feb 2025 that he intends to impose "new 25% or higher" tariffs on a range of product imports including semiconductors, possibly on 2 April 2025. Bloomberg Bloomberg: 18 Feb 2025

Donald Trump's New and Increased Tariffs on China, Canada and Mexico.

Update 28 Feb 2025: Tariffs on Canada and Mexico commence on 4 Mar 2025 and an additional 10% tariff will be imposed on China's imports to USA.

White House announcement 20 Jan 2025. Press coverage:

• 🌎 Global tariffs: BBC 1 Feb 2025
• 🌎 Reciprocal tariffs: The Times 7 Feb 2025

⏯️ Trump has applied 25% tariffs on imports from USA's neighbours, 🇲🇽 Mexico and 🇨🇦 Canada, and an additional 10% tariff on imports from 🇨🇳 China now 35%) Saturday 1 Feb 2025: Bloomberg 31 Jan 2025. Together, 🇨🇳 China, 🇲🇽 Mexico and 🇨🇦 Canada accounted for more than 40% of imports into 🇺🇸 USA BBC 2 Feb 2025.

🇨🇳 The Chinese products that are now subject to the 35% tariff are:

• Electronics and Technology This includes items like smartphones, computers, and semiconductors, as well as parts and components used in tech manufacturing.
• Automotive Parts Components for vehicles: tires, batteries, electrical systems used in cars and trucks; home goods and furnishings, including furniture made from wood, plastic, and metal.
• Medical and Pharmaceutical Products Medical equipment, devices, and some pharmaceutical products were included as well.
• Machinery and Equipment Industrial machinery, engines, and electrical equipment that are essential for manufacturing and construction.
• Consumer Goods Clothing, footwear, various household items, toys, games, and sports equipment.

Reuters reports there's nothing Canada, China or Mexico can do to delay the implementation of tariffs on 1 Feb 2025.

China Responds to USA's Tariffs With Levies

China is expected to impose 10%-15% levies on US energy and agricultural goods. Bloomberg 4 Feb 2025

Tariffs on imports from Canada and Mexico delayed until Mar 2025 ⏯️ Bloomberg 13 Feb 2025.

Trade Restrictions between China and USA

• 🇺🇸 USA considers restrictions on flow of advanced chips to China
• 🇨🇳 China considers restrictions on dumping of mature chips from USA

🇨🇳 China's Ministry of Commerce (MOFCOM) has added 28 U.S. companies to its "Export Control List" and 4 U.S. companies to its "Unreliable Entities List" "for their selling arms to China's Taiwan region" China Daily 15 Jan 2025, Reuters 15 Jan 2025

🇨🇳 MOFCOM's "Unreliable Entities List" was introduced in 2019 as part of a broader set of measures to counter foreign entities and individuals it deems to be harming Chinese interests. This list includes foreign companies, organizations, and individuals who China believes are engaged in activities that undermine China's sovereignty, security, or development. China Daily

🇨🇳 It seems that MOFCOM's list is intended to identify and obstruct foreign entities and individuals it deems to be harming Chinese interests. This list includes foreign companies, organizations, and individuals who China believes are engaged in activities that undermine China's sovereignty (eg of Taiwan), security, or development. It maintains an "Export Control List", which includes some 20 companies that are blocked from exports of dual-use items.

🇺🇸 US companies that have faced scrutiny from China in the past and may be under consideration or already listed include (sources include ChatGPT):

• Boeing is another US company in the defense and aerospace sector that has been scrutinized for its dealings with China, especially over technology transfers and intellectual property concerns. While it is more prominent in aerospace, its technological ties have led to occasional tensions.
• Cisco: Cisco has been caught in the crossfire of US-China tech disputes. As a leading networking technology provider, it has faced accusations of enabling surveillance and monitoring tools in China, which the government has used for internal security purposes.
• General Atomics Aeronautical Systems, Inc (unmanned aircraft)
• General Dynamics Land Systems
• ✅ Intel: Intel has faced restrictions and limitations in its business dealings in China due to tensions over technology exports and intellectual property issues. There have also been instances where China has criticized US tech giants like Intel for contributing to technological decoupling efforts.
• Lockheed Martin: While primarily a defense contractor, Lockheed Martin has been involved in technology development related to military applications. China's criticism of US arms sales to Taiwan and other geopolitical issues likely contributed to its inclusion on the list.
• Micron Technology (added in May 2023)
• Qualcomm: Qualcomm has faced scrutiny over its dealings with Chinese telecommunications companies. In particular, the company has faced issues regarding its licensing agreements and its perceived influence over Chinese technology firms. Qualcomm was also affected by broader tensions between the US and China over the trade war.
• RTX Corporation (formerly Raytheon Technologies): Like Lockheed Martin, Raytheon is a major US defense contractor involved in the development and production of defense-related technology, including missile systems. The company has been implicated in US arms sales to Taiwan and other regions that China sees as problematic.

The only company included in our portfolio that is directly affected by MOFCOM's policies is ✅ Intel. However collateral impacts may be felt by ✅ TSMC and NASDAQ-listed GPU companies whose supply chains may be impacted by MOFCOM's policies.

TikTok

In addition to the examples above is the US Supreme Court (SCOTUS) order: "As of January 19, the Protecting Americans from Foreign Adversary Controlled Applications Act will make it unlawful for companies in the United States to provide services to distribute, maintain, or update the social media platform TikTok, unless U. S. operation of the platform is severed from Chinese control." In other words SCOTUS ordered that TikTok be shut down for US customers or sold. On 17 Nov 2024 TikTok parent ByteDance valued itself at $300 billion in a buyback offer.

TikTok ceased to be available in USA on 17 Jan 2025. However on 19 Jan 2025 President-Elect Trump said he would aim to reinstate TikTok to allow the company more time to find a buyer. TikTok defied the Congress bill and, on the basis of Trump's declaration, restored the service to US users. CNN 19 Jan 2025 Trump has previously said that Elon Musk [X] or Larry Ellison [Oracle] would be eligible buyers of TikTok, however the owner of ByteDance has opposed the sale (to any buyers). Bloomberg has recently reported that Chinese officials were considering the possibility of selling TikTok’s U.S. operations to Musk [NBC 22 Jan 2025].

An alternative may be expressed interest from U.S. company PerplexityAI:

• PerplexityAI focuses on advanced search and answering questions, with a research-oriented foundation
• JasperAI focuses on content creation with heavy backing from venture capital. In December 2024 Perplexity was valued at USD 9 billion.

⏏️

🇺🇸 Assurance of Electrical Power for Both AI and Households

AI data centers are multiplying across the US and sucking up huge amounts of power. New evidence shows they may also be distorting the normal flow of electricity for millions of Americans Bloomberg: 28 Dec 2024 Bloomberg: 13 Dec 2024

✅ Amazon bought more renewable power in 2024 than any other company. As part of it Climate Pledge commitment to reach net-zero carbon emissions by 2040, Amazon is supporting renewable energy projects on a global scale. For the fifth year in a row, Amazon is the largest corporate purchaser of renewable energy globally, according to Bloomberg NEF, and supporting more than 600 wind and solar projects. Together these projects will be capable of powering the equivalent of 8.3 million U.S. homes. Amazon 17 Jan 2025

US President Joe Biden signed an Executive Order on 14 Jan 2025 " directing the Department of Defense and the Department of Energy to lease federal sites where the private sector can build frontier AI infrastructure at speed and scale. These efforts are designed to accelerate the clean energy transition in a way that is responsible and respectful to local communities, and in a way that does not impose any new costs on American families."

This initiative is intended to "secure American leadership in artificial intelligence (AI) and ensure that the infrastructure needed for advanced AI operations —including large-scale data centers and new clean power infrastructure— can be built with speed and scale here in the United States."

White House, "Statement by President Biden on the Executive Order on Advancing U.S. Leadership in Artificial Intelligence Infrastructure" 14 Jan 2025

⏏️

Stargate AI Project

President Donald Trump on 21 Jan 2025 announced a new USD 500 billion, private sector investment to build artificial intelligence infrastructure in the US, with Oracle, ChatGPT creator OpenAI, and Japanese conglomerate SoftBank among those committing to the joint venture. ⏯️ Trump, CNN 21 Jan 2024

The joint venture, called Stargate, is expected to begin with a data center project in Texas, according to CBS News

• Masayoshi Son, founder, CEO and largest shareholder of SoftBank will be Stargate’s chairman. SoftBank will have financial responsibility for Stargate.
• OpenAI will have operational responsibility.
• Oracle may shift some AI training workloads from ✅ Microsoft Azure to Oracle Cloud.
• MGX, a fourth partner, is also contributing funding.

Company execs were expected to commit to an initial investment of USD 100 billion at an appearance at the White House Tuesday. Other companies are expected to join the venture and bring investment in the program up to USD 500 billion in the coming years.

Oracle stock jumped 7% on 21 Jan 2025 and continued its ascent in pre-market trading on 22 Jan 2025, rising over 8%.

Elon Musk has cast doubts on the Stargate members' ability to pull together funding necessary for the project:

• Elon Musk post on X 1535hr 22 Jan 2025: “They don’t actually have the money” Musk wrote in a post on X. “SoftBank has well under $10 billion secured. I have that on good authority.”
• Sam Altman reply on X 0342hr 28 Jan 2025: "wrong, as you surely know. want to come and visit the first site already underway? this is great for the country. i realise what is great for the country isn't always optimal for your companies, but in your new role I hope you put 🇺🇸 first."

⏯️ ElonMusk, CNN 22 Jan 2024 Elon Musk was a co-founder of OpenAI in 2015, but not a founding investor of a specific AI product or startup. He was one of the early significant backers with a commitment in the $1 billion range for OpenAI.

Bloomberg 24 Jan 2025: ⏯️ "Musk, Altman Spar Over Stargate and Trump"

⏏️

BCI: Synchron's Chiral: Cognitive AI Brain Foundation Model

"Synchron, a category-defining brain-computer interface (BCI) company, today unveiled the roadmap to Chiral™, a foundation model of human cognitions, marking the emergence of Cognitive AI—artificial intelligence trained directly on human neural activity. Synchron is advancing BCI from supervised to self-supervised learning, accelerating the transition by combining large-scale neural data with advanced NVIDIA AI-powered computing." Business Wire 19 Mar 2025

⏏️

France Reports EU109 billion (USD112 billion) in Private Investments in the AI Ecosystem

On 9 Feb 2025 French president, Emmanuel Macron, reported a total of EUR 109 billion in private investments in the AI ecosystem — or around USD 112 billion at current exchange rates. TechCrunch.

⏏️

Apple will spend more than $500 billion in USA over the next 4 years

Apple 24 Jan 2025: ⏯️ Apple will spend more than $500 billion in the U.S. over the next four years

⏏️

🇯🇵 SB OpenAI Japan: OpenAI and Softbank JV for Japanese Business AI

🇯🇵 SoftBank Group CEO Masayoshi Son said he and OpenAI CEO Sam Altman have agreed to set up SB Open AI Japan, a joint venture to offer AI services to corporate customers. Softbank will pay $3 billion annually to use OpenAI's technology across SoftBank group companies. SoftBank is also considering investing USD 15 billion to USD 25 billion in the company. Reuters 3 Feb 2025 AFR 30 Jan 2025

⏏️

🇨🇳 Apple iPhone Sales Underperform in China

Analysts report underperformance of ✅ Apple's iPhone sales is attributable to Apple Intelligence. Market shares: Vivo 17%, Huawei 16%, Apple 15% ⏯️ CNBC: 17 Dec 2024

⏏️

ECONOMIC OUTLOOK: 2025, 2026

🌎 "Global growth is projected at 3.3% both in 2025 and 2026, below the historical (2000–19) average of 3.7%. The forecast for 2025 is broadly unchanged from that in the Oct 2024 World Economic Outlook (WEO), primarily on account of an upward revision in the United States offsetting downward revisions in other major economies. Global headline inflation is expected to decline to 4.2% in 2025 and to 3.5% in 2026, converging back to target earlier in advanced economies than in emerging market and developing economies."

Other reports:

• 🇦🇺 ⏯️ PBO: Australian Fiscal Outlook: 29 Jan 2025
• 🌎 ⏯️ IMF: World Economic Outlook: Current and Previous Issues
• 🌎 ⏯️ IMF: Global Financial Stability Report: 22 Oct 2024
• 🌎 ⏯️ IMF: Global Financial Stability Report: 22 Oct 2024: Global Debt/GDP: 22 Oct 2024
• 🌏 ⏯️ IMF: Regional Economic Outlook for Asia and Pacific, November 2024 | Resilient Growth but Higher Risks: 31 Oct 2024
• 🇺🇸 US Household Debt Delinquency Hits the Highest in Almost 5 Years The Federal Reserve Bank of New York’s Center for Microeconomic Data Quarterly Report on Household Debt and Credit shows total household debt increased by $93 billion (0.5%) in Q4 2024, to $18.04 trillion.
  Household debt balances continue steady increase; Delinquency transition rates remain elevated for auto and credit cards ⏯️ FRBNY: 13 Feb 2025

⏏️

🇦🇺 AUSTRALIAN FISCAL OUTLOOK 2024-2025

🇦🇺 Australia's 2024-2025 National fiscal outlook includes a forecast national net debt that is estimated to reach 36% of GDP by 2027‑28. One of several consequences of this is an increasing burden of debt interest payments (revised up from $48.1 billion by $4.5 billion for 2026‑27, to $52.6 billion) which add to the government's overspending and difficulty in seeing a floor to Australia's debt.

The impact of increased debt and probable devaluation of the AUD on US investment portfolio is positive. We'd rather see some competent fiscal management, but we're not expecting to see this in the forseeable future as failure to manage increases the debt and the difficulty of reaching a turning point.

🇺🇸 On the other side of the Pacific, USA is also struggling with its fiscal deficits and debt. President Trump's main problem is that he does not have the power to control the US administration's fiscal programs. For example Trump tried to “temporarily pause all activities related to obligation or disbursement of all Federal financial assistance ... to provide the Administration time to review agency programs and determine the best uses of the funding for those programs consistent with the law and the President’s priorities", but found that he does not have that power. In the meantime, USA's debt and interest continue to rise. As at 27 Jan 2025 USA's public debt is USD 36.219 trillion, 123% of USA's GDP in Q3/2025 (USD 29.37 trillion).

⏏️

External Papers

Bloomberg: AI Glossary

Quantum Insider: Market Intelligence White Paper [Jan 2025]

World Economic Forum (Davos 2025) and McKinsey: Market Intelligence White Paper [Jan 2025]

McKinsey Digital: Technology Trends Outlook 2024

⏏️

RETROSPECT: EVOLUTION OF WALL STREET TRADING

Evolution of Wall Street: Behind The Ticker Tape

Evolution of Wall Street: Evolution Of The Ticker Tape

Wall Street Today: New York Stock Exchange