About TechInvestment.com

Our Investment Portfolio

TechInvestment.com [Bunting Pty Ltd: ABN 64 096 255 668] is a long equity investor founded 25 years ago in March 2001. It invests in NASDAQ, NYSE and ASX listed technology companies.

Our focus is on advanced technologies, including:

• 🇺🇸 Artificial Intelligence (AI) via ✅ Palantir, ✅ Google, ✅ Meta and others,
• 🇺🇸 Graphics Processing Units (incl GPU, TPU, LPU) via ✅ Nvidia and previously AMD [MI455X]. Nvidia's quarterly profit hit USD 43 billion on AI chip sales. Total sales for the fiscal year were USD 130 billion [NY Times] - Update . We have previously been an investor in American Micro Devices [AMD]. [Update: 18 Feb 2026]
• 🇺🇸 Tensor Processing Units (TPU) via ✅ Google [for use with tensor flow applications],
• 🇺🇸 Language Processing Units (LPU) [developed by Groq],
• 🇺🇸 Quantum Computing (QC) via ✅ D-Wave, ✅ Rigetti, and ✅ Amazon via AWS Braket. We have previously invested in IonQ [IONQ]
• 🇺🇸 Data Centers in USA and Australia, via 🇦🇺 ✅ NextDC, 🇺🇸 ✅ Constellation and others,
• 🇺🇸 nuclear, coal, gas, wind and other forms of power generation for data centers, and
• 🇺🇸 humanoid social robotics via ✅ Tesla's Optimus Gen 3 robots and Boston Dynamics. We have previously invested in [Intuitive Surgical ISRG]
• ⏯️ Manufacturing ✅ Tesla Optimus humanoid robots. More on social robotics.

In 2025 sovereign wealth and public pension fund investors poured USD 132 billion, roughly half of their total investments for the year, into USA.

Sovereign wealth funds managed USD 66 billion in AI and digitisation in 2025, according to the "Global SWF Year in Review 2025: The Ideas, Power Shifts and Structures That Defined Sovereign Capital", 30 Dec 2025.

US economic, trade, political and international issues (especially USA's military actions in the Middle East in 2026) create a great deal of uncertainty (risk) in US investment. This also increases AUD:USD currency risk - which may represent an additional cost to our AUD-denominated portfolio.

Global Election Calendar 2026

The most significant elections in 2026 (with potential impacts on economic or investment environments) are:

• 13 Sep 2026: 🇪🇺 Europe (🇸🇪 Sweden)
• Sep 2026: 🇷🇺 Russia
• 3 Nov 2026: 🇺🇸 USA Mid Term Election
• Before 19 Dec 2026: 🇳🇿 New Zealand

🌎 Portfolio Performance Issues and News

Artificial Intelligence

28 Mar 2026: Jensen Huang, Nvidia CEO ⏯️ discusses the limits of AI scaling laws with Lex Fridman

20 Mar 2026: ✅ Palantir's Maven artificial intelligence system will become an official program of record, Deputy Secretary of Defense Steve Feinberg said in a letter to Pentagon leaders, a move that locks in long-term use of Palantir’s weapons-targeting technology across the U.S. military. In the 9 Mar 2026 letter to senior Pentagon leaders and U.S. military commanders, Feinberg said embedding Palantir’s Maven Smart System would provide warfighters “with the latest tools necessary to detect, deter, and dominate our adversaries in all domains”.

✅ Palantir's Maven will be hosted out of ✅ Amazon Web Services (AWS)’s Stockholm region. Medium

16 Jan 2026: ✅ Anthropic (Claude AI) to open an office in 🇦🇺 Australia

13 Jan 2026: ✅ Meta cuts 10% of employees in its Reality Labs (VR Metaverse) division

13 Jan 2026: ✅ Meta unveils 'Meta Compute' initiative to build AI infrastructure

13 Jan 2026: ✅ Meta pivots from Metaverse to AI devices such as RayBan Meta AI glasses.

12 Jan 2026: Bloomberg: ✅ Apple and ✅ Google confirm a multi-year deal for Gemini to run AI-powered Siri later in 2026.. Barron's reports that ✅ Google is now the top AI company as ✅ Alphabet stocks rise past USD 4 trillion.

10 Jan 2026: ✅ Meta signs multi-gigawatt nuclear deals to power AI

23 Jul 2025: 🇺🇸 White House AI Action Plan

26 Jun 2025: AI’s Consumer Tipping Point Has Arrived: "More than half of American adults (61%) have used AI in the past six months, and nearly one in five rely on it every day. Scaled globally, that translates to 1.7–1.8 billion people who have used AI tools, with 500–600 million engaging daily. This is no longer experimentation; it’s habit formation at an unprecedented scale." [MenloVC]

14 Apr 2025: The NATO Communications and Information Agency (NCIA) and ✅ Palantir Technologies finalized the acquisition of the Palantir Maven Smart System NATO (MSS NATO) for employment within NATO’s Allied Command Operations (ACO), marking a significant advancement in the modernization of NATO’s warfighting capabilities." [NATO Supreme Headquarters Allied Powers Europe (SHAPE): 14 Apr 2025]

Quantum Computing

ChatGPT comment: "The strange part isn’t that quantum computing is hard. It’s that we ever expected it to be otherwise."

22 Jan 2026: 🇦🇺 Australia – 🇯🇵 Japan Quantum Computing Partnership launches on international stage. The agreement focuses on integrating quantum computing with supercomputing to accelerate scientific breakthroughs at the annual supercomputing conference in the region. Australia’s CSIRO, on behalf of the Pawsey Supercomputing Research Centre, has signed a Letter of Intent with Japan’s National Institute of Advanced Industrial Science and Technology (AIST) to strengthen scientific cooperation in quantum computing.

In 2024 IBM and PsiQuantum were awarded DARPA US2QC grants to develop a fully fault-tolerant quantum computer with logical qubits protected by quantum error correction (primarily surface-code-style schemes) for its Validation and Co-Design stage of the Underexplored Systems for Utility-Scale Quantum Computing (US2QC) program , one of two programs that make up DARPA’s larger Quantum Benchmarking Initiative (QBI). The goal of the QBI is to see if an industrially useful quantum computer can be built by 2033.

Companies selected for Stage B include:

• Atom Computing: Boulder, Colorado (scalable arrays of neutral atoms)
• Diraq: Sydney, Australia, with operations in Palo Alto, California, and Boston, Massachusetts (silicon CMOS spin qubits)
• IBM - Thomas J Watson Research Center: Yorktown Heights, New York (quantum computing with modular superconducting processors)
• Nord Quantique: Sherbrooke, Quebec, Canada (superconducting qubits with bosonic error correction)
• Photonic Inc: Vancouver, British Columbia, Canada (optically-linked silicon spin qubits)
• Quantinuum: Broomfield, Colorado (trapped-ion quantum charged coupled device (QCCD) architecture)
• Quantum Motion: London, UK (MOS-based silicon spin qubits)
• QuEra Computing: Boston, Massachusetts (neutral atom qubits)
• Silicon Quantum Computing Pty Ltd: Sydney, Australia (precision atom qubits in silicon)
• Xanadu: Toronto, Canada (photonic quantum computing)

Semiconductors

26 Mar 2026: ✅ New Coreweave Chip.

25 Jan 2026: ✅ Nvidia has begun building a Tesla Giga semiconductor chip design plant (Terafab) in Austin Texas. Elon Musk has said Tesla and SpaceX aim to build what could become the world’s largest chip manufacturing facility, potentially spanning tens of millions of square feet and supporting artificial intelligence and space-based data systems. [Austin-American Statesman 26 Mar 2026].

26 Jan 2026: Microsoft takes aim at ✅ Google, ✅ Amazon, and ✅ Nvidia with new Maia 200 AI chips [Yahoo Finance].

26 Jan 2026: Microsoft has announced the launch of its latest chip, the Maia 200, which the company describes as a silicon workhorse designed for scaling AI inference. The Maia 200, which follows the company’s Maia 100 released in 2023, has been technically outfitted to run powerful AI models at faster speeds and with more efficiency. Maia 200 comes equipped with over 100 billion transistors, delivering over 10 petaflops in 4-bit precision and approximately 5 petaflops of 8-bit performance — a substantial increase over its predecessor.

Inference refers to the computing process of running a model, in contrast with the compute required to train it. As AI companies mature, inference costs have become an increasingly important part of their overall operating cost, leading to renewed interest in ways to optimize the process.[TechCrunch 26 Jan 2026]

16 Jan 2026: Hardware firms hit by rising memory chip costs. [Bloomberg]

15 Jan 2026: TSMC, the world's largest contract chipmaker, posted a Q4/2025 record profit: up 35% to USD 16.01 billion [Reuters] TSMC Forecast Lifts Peers on Robust AI Demand [Bloomberg]

14 Jan 2026: Washington permits ✅ Nvidia to sell its H200 chip to China.

8 Jan 2026: US Commerce Department approves ✅ Nvidia H200 AI chip sales to China.

8 Jan 2026: China plans to approve some imports of ✅ Nvidia’s H200 chips as soon as this quarter. [Bloomberg] "This move represents a major win for Nvidia. China is the world’s largest market for semiconductors, and CEO Jensen Huang has said that the AI chip segment alone could generate USD 50 billion in the coming years." About the Nvidia H200 chip.

30 Dec 2025: High-Numerical Aperture (High-NA) EUV Production: ✅ Intel [Behind Asia 30 Dec 2025].

High-Numerical Aperture (High-NA) EUV AI Chip Timeline

2018–2021 (R&D phase) — ASML + Intel + partners:

• Joint development of next-generation extreme ultraviolet (EUV) tools (0.55 NA [NA: High Numerical Aperture]) to extend Moore’s Law beyond conventional EUV (0.33 NA).
• High-NA roadmap targets production readiness ~2025.

2022 — Intel + ASML:

• Formal collaboration announced to bring High-Numerical Aperture (High-NA) EUV into manufacturing.
• Introduction of TWINSCAN EXE platform (EXE:5000 / EXE:5200) for sub-2 nm chips.

2023–early 2024 — Intel (first adopter):

• Intel receives the first High-Numerical Aperture (High-NA) EUV system from ASML for installation and testing.

April 2024 — ASML (+ Intel, others):

• Shipment of additional High-NA systems to leading chipmakers (Intel, likely Samsung/TSMC).
• Tools priced ~$350M–$400M each; still primarily R&D / pilot use.

2025 (early deployment & testing) — Intel:

• Installs multiple High-NA machines; processes tens of thousands of wafers in development environments.

2025 (industry transition) — ASML + ecosystem:

• High-Numerical Aperture (High-NA) EUV recognized as critical for future AI chips (≈1.4 nm and below).
• Enables ~8 nm resolution vs ~13 nm for previous EUV.

Late 2025 — Intel (first production milestone):

• Achieves high-volume manufacturing (~30,000 wafers/quarter) using High-Numerical Aperture (High-NA) EUV for 18A (~1.8 nm) node.
• Marks first real use in advanced chip production (including AI processors).

2026 — ASML + Intel + Samsung:

• High-Numerical Aperture (High-NA) EUV systems declared production-ready, with proven wafer throughput and uptime.
• Intel installs first commercial High-NA tool (EXE:5200B) for next-gen nodes.

2026–2027 (risk production phase) — Intel, Samsung, SK hynix:

• Companies begin risk production and early node integration using High-NA EUV.
• Technology tested for AI chips at sub-2 nm nodes.

2027–2028 (expected mass adoption) — Intel, Samsung:

• Start of mass production using High-Numerical Aperture (High-NA) EUV for ~1.4 nm logic and advanced memory.

~2028–2030 (selective adoption) — TSMC:

• Continues using conventional High-Numerical Aperture (High-NA) EUV for early 1.4 nm nodes.
• Delayed High-Numerical Aperture adoption (≈2030+) due to cost and process tradeoffs.

Current Status

• ASML is the sole supplier enabling High-Numerical Aperture (critical bottleneck technology).
• Intel is the first mover, already using High-NA in early production for AI-class nodes.
• Samsung & SK hynix follow closely for memory and logic.
• TSMC is a fast follower, prioritizing cost efficiency over early adoption.
• High-Numerical Aperture is essential for future AI chips (<2 nm) due to improved resolution and reduced multi-patterning.

Architectures

1. H100 (Hopper architecture) — Baseline (2022–2023)

What it introduced:

• 4th-gen Tensor Cores + Transformer Engine (FP8 support)
• 80 GB HBM3, ~3.35 TB/s bandwidth
• ~80B transistors

Impact:

• First GPU truly optimized for LLMs + generative AI at scale
• Became the industry standard for training models up to ~70B parameters

2. H200 (Enhanced Hopper) — Memory-driven upgrade (2024)

Key improvement: memory, not compute

• 141 GB HBM3e (+76% capacity)
• 4.8 TB/s bandwidth (+43%)
• Same compute silicon as H100

What changed:

• No major increase in raw FLOPS
• Massive improvement in memory-bound workloads

AI impact:

• Larger models fit on a single GPU (100B+ params)
• Longer context windows
• Higher inference throughput

Step change type:

• Scale-up improvement (bigger models, not faster compute)

Nvidia Blackwell:

• B100 (early Blackwell)
• Nvidia GB300 NVL72
• NVIDIA HGX B300

Energy

19 Mar 2026: ✅ Constellation Energy (CEG) will sell 4.4GW of PJM assets to LS Power for $5B as part of the Calpine acquisition. [Power Engineering]

Constellation Energy has in recent months signed long-term power purchase agreements with Microsoft, ✅ Meta and CyrusOne, while moving to acquire Calpine and restart nuclear capacity such as Three Mile Island under supportive US federal policies.

9 Feb 2026: ✅ Constellation Energy (CEG) announced that Calpine LLC, a business unit of Constellation Energy, signed a new 380-megawatt (MW) agreement with Dallas-based CyrusOne [CONE], a leading global data center developer and operator, to connect and serve a new data center adjacent to the Freestone Energy Center, in Freestone County,Texas. The agreement provides CyrusOne with access to power, grid connectivity and site infrastructure needed to support development of the new facility, while ensuring electricity continues to flow to the regional grid and ensuring reliability for all customers and communities. Calpine has also entered into an exclusive agreement to provide power, grid connectivity and site infrastructure for Phase 2, which will be an additional 380 MW. These agreements are in addition to the 400 MW agreements announced in the second half of last year between Calpine and CyrusOne for the Thad Hill Energy Center in Bosque County, Texas. [CyrusOne]

16 Jan 2026: President Trump and the governors of several US Northeastern states agreed to push for an emergency wholesale electricity auction that would compel technology companies to effectively fund new power plants [Bloomberg].

Techinvestment.com is an investor in ✅ Constellation Energy (CEG), which has 55 gigawatts of capacity from nuclear, natural gas, geothermal, hydro, wind and solar facilities. ✅ Constellation partners with Calpine LLC, America's largest generator of electricity from natural gas and geothermal resources with operations in competitive power markets. The fleet of 79 energy facilities in operation represents over 27,000 MW of generation capacity.

Data Centers

2 Feb 2026: More than USD 3 trillion is expected to be spent building the data centers needed for the AI boom, and much of the funding is coming from debt markets [Bloomberg].

🇺🇸 Potential US Political Crisis

US political uncertainty/chaos and policies (eg tariff uncertainty, alienation of allies, and US-initiated wars) also have an impact on valuation of stocks, so we're interested in the very unstable performance of US politics and politicians,
especially when the chickens come home to roost.

🇺🇸 There are more than 230 current litigation actions involving Donald Trump and his administration as at 15 Mar 2026.

It is reported that USA may, in the near term, face a political crisis as Congress proposes impeachment and removal of Donald Trump as President.

• ⏯️ Warren Buffett discusses impeachment and removal of a president: 15 January 2026
• ⏯️ Video report: CBS News, 15 January 2026
• ⏯️ Video report: NBC News Special, 15 January 2026
• 🇺🇸 ⏯️ Video report: Senate Judiciary Committee, during a hearing for judicial nominees, questions nominees' reaction to political pressure in their future work as judges: 25 Mar 2026

🇺🇸 Trump seems to have got back to his "The Apprentice" days when he fired candidates every episode ... except that the people he fires as President are senior office holders in his government. And he seems to think that military conflict will earn support - maybe missing the point that such support only happens when USA is defending its interests, not attacking other countries' interests.

Trivia:

• In Donald Trump's first term as President, White House staff turnover was very high [Brookings Institution: 19 Jan 2018]
• Donald Trump's "The Apprentice" ran for about 194 episodes in 15 seasons. Trump hired 6 apprentices with roles in Trump Tower (1), Trump brand liaison (1), Trump organisation (1), project management (2). [Source: OpenAI's ChatGPT]