ICM HPQC News Flash - December 2025
- ICM
- 2 days ago
- 9 min read
ICM HPQC Fund (“ICM HPQC”) continues to see excitement in the next-generation computing space. In this newsflash, we cover:
A look back at an exceptional year for compute and AI infrastructure:
Market trends shaping hyperscaler investment, data-centre expansion and energy constraints
The rise of agentic AI, productivity tools and record valuations across the sector
A breakout year for quantum computing, government initiatives and strategic momentum
How these shifts position ICM HPQC for 2026 and beyond
Portfolio updates across Mixx Technologies, Diraq, Q-CTRL and Salience Labs
A curated roundup of notable industry news
Year in Review
As we come to the end of the year, it’s a good moment to reflect on what has been an extraordinary year in compute.
At a fund level, we’ve made four investments - Salience Labs, Diraq, Q-CTRL and Mixx Technologies. Throughout the year, we met and analysed hundreds of companies, attended more than 20 international conferences, and continued to sharpen our thesis on what the next generation of compute will look like. We are now approaching our final close and are already preparing to hit the ground running in 2026, where our focus remains simple: find and back the best teams solving real bottlenecks in compute.
The momentum in the market has exceeded even our optimistic expectations. 2025 was the year investments in compute technologies truly broke out. Hyperscalers and the
leading AI companies now treat compute as a core competitive advantage - arguably the competitive advantage. We saw unprecedented growth in data centre buildouts, both in the size of capital commitments and in the physical footprint of new sites. Meta plans to build a Manhattan-sized data centre requiring the equivalent of 5 nuclear reactors to operate. Total global data centre capex spend from Amazon, Microsoft, Alphabet and Meta increased 69% from US$218B in 2024 to US$369B in 2025, almost tripling the historic CAGR (figures based on Company results and announcements). These are numbers normally associated with national-scale infrastructure, not individual technology verticals.
Meanwhile, AI’s value has firmly moved beyond chatbots. New startups in productivity, healthcare and enterprise tooling are now reshaping day-to-day work. Apps like Fireflies and Granola have become ubiquitous in Teams and Zoom calls—automating note-taking to the point where many people will never manually write minutes again. “Vibe coding” tools like Loveable, Cursor and Claude Code let anyone turn ideas into prototypes at remarkable speed, shrinking development cycles from weeks to hours.
2025 has also seen the shift to agentic AI, where AI is given the freedom to make its own decisions as it relates to certain workflows or products. In less than a year, reasoning-optimised models have increased from negligible to over 50% of all AI token volume. During that time, average prompt lengths have quadrupled to over 6,000 tokens. Exponential View sees this as the “mechanical footprint of agentic loops iterating in the background”. We expect this trend toward agentic products to continue, reflected by a shift to inference-optimised hardware.
This isn’t just anecdotal: Anthropic estimates that current-generation AI models could lift annual US labour-productivity growth by 1.8% over the next decade - effectively doubling the country’s average annual productivity growth since 2019. Unsurprisingly, this excitement has translated into towering valuations: NVIDIA became the first company in history to cross both the US$4 trillion and US$5 trillion milestones, OpenAI’s “cash-burning machine” is valued around US$500B, and OpenAI-competitor Thinking Machines closed a seed round at a US$12B valuation. The market is excessively exuberant, but the underlying demand is very real.
At the same time, there’s been an increasing focus on the very real pressures on water and
energy. The FT’s latest article visualises the power crunch side of this well. Data centres in America represent a combined capacity of 51GW (5% of peak demand). An estimated 44GW of additional capacity will be required by 2028. Source Material’s exposé on water consumption showed that tech companies are building data centres in water-scarce regions on five continents.
Quantum computing also had a watershed year. Nvidia's Jensen Huang shifted from saying quantum computing was “15–30 years away” to launching the DGX Quantum programme, building its quantum ecosystem and investing in QuEra, Quantinuum and PsiQuantum through Nvidia's venture arm NVentures. The US government’s DARPA announced both Stage A and Stage B selections for its Quantum Benchmarking Initiative, identifying 11 companies globally with credible roadmaps to practical quantum advantage by 2033, following a 6 month evaluation process. We also saw the first major acquisition in this sector: IonQ’s purchase of Oxford Ionics for US$1B. Valuations are now entering the decacorn era with multiple companies worth over US$10B. In November, Google’s CEO joined a long line of tech-CEO optimism for quantum saying “I would say quantum is where AI was five years ago. So, I think, in five years from now, we'll be going through a very exciting phase in quantum.” The quantum market is incredibly hot - likely too hot - but it reflects a growing belief that quantum will meaningfully reshape compute rather than remaining a distant research project.
November brought another major development: the US government’s Genesis Mission - an “Apollo Programme for Compute.” Its goal is to create a national AI-enabled science and discovery engine, combining high-performance computing, AI models, federal datasets and autonomous “AI agents” capable of running experiments, testing hypotheses and automating research. This is an explicit national-level acknowledgment that scientific progress is now bottlenecked by compute. It’s another sign of the US’ commitment to dominance in the AI space. Director of the Genesis Missions Dario Gill considers both AI and quantum computing as the key drivers in the new revolution in computing.
All of this positions ICM HPQC extremely well. We’re operating at the centre of a structural jump in demand for AI and specialised compute, where “increasing compute is
the literal key to increasing revenue.” Massive investment into data-centre infrastructure creates space for entirely new approaches, while physical and energy bottlenecks force innovation rather than incrementalism. Demand is broad, global and tied to long-run productivity growth.
Our portfolio companies are directly targeting these bottlenecks:
Salience Labs is solving data-connectivity limits that slow down AI systems.
Diraq - both an Nvidia partner and a DARPA QBI participant - is building a credible path to quantum computing co-located with data centres.
Q-CTRL is making quantum computing useful sooner by stabilising and optimising the underlying hardware.
Mixx Technologies is going after the holy grail in compute: integrating GPUs and optics on the same package to overcome power and bandwidth ceilings.
2025 showed that compute is one of the most important investment themes globally, and we’re well-positioned for what comes next.
Portfolio News
Mixx Technologies
In November, we welcomed Mixx Technologies to the portfolio. We are exceptionally excited about this investment and feel privileged to join industry veterans turned co-founders Dr Rebecca Schaevitz and Vivek Raghuraman on this journey.
TDK Ventures, Applied Ventures, G Vision Capital and other strategics joined us for this investment, highlighting the conviction of investors who live and breathe this space.
Mixx Technologies is working on silicon-integrated optics (or co-packaged optics “CPO”), which will be the backbone of the next generation data centres, enabling data to be moved and processed at higher speeds with greater energy efficiency than existing solutions.
Rebecca and Vivek were the innovators behind Intel’s silicon photonics-based transceivers and Broadcom’s industry-first co-packaged optics (CPO) for network switches and are now teaming up to build the next generation of data centre connectivity. We believe they have the “right to win” in this space.
As mentioned in last month’s newsflash, CPO is a massively important part of the AI data centre buildout which is seeing unprecedented investment from the hyperscalers. Analyst firm LightCounting expects CPO to start taking hold of the market in 2026 and 2027 with major growth expected from 2029.
In the same week, Celestial AI was applied by Marvell for up to $5.5B (contingent on meeting certain milestones), highlighting the appetite and possible outcomes for photonic connectivity companies. Read the press release here.
CEO & Founder Andrew Dzurak was awarded the 2025 Innovation Leadership Award by Innovation Australia, with judges praising his “rare combination of scientific insight, commercial vision and national-interest leadership”, noting that his work positions Australia as a genuine contender in the global quantum race. Read here.
Q-CTRL reflects on 2025 with its year in review. Highlights included:
Time – Best Inventions of 2025
>$50M of contract revenue secured
100x quantum advantage achieved in navigation
Q-CTRL also announced the launch of the Quantum Utility Block, a family of full-stack, pre-validated quantum computer reference implementations, alongside QuantWare and Qblox. It is designed to make it easier to procure quantum systems or on-premises deployments thereby making full-scale quantum systems useful, accessible, and economically viable to a range of users. Read more.
Q-CTRL is now integrating its Fire Opal product with an IBM Quantum computer based at RIKEN in Japan. Read more.
Salience Labs (Salience) has added additional investors to its cap table including the Oman Investment Authority (OIA) – a US$53B sovereign wealth fund.
Salience’s team is growing fast and they’ve outgrown their Oxford Lab. In November, Salience moved into a 6,700 sq ft R&D facility at Milton Park. Read here.
Salience was named in the Sifted AI 100 at #28 – a ranking of Europe’s most influential startups. Read here.
What We're Reading
Akamai introduced Inference Cloud, leveraging edge infrastructure and Nvidia GPUs to address AI deployment challenges by bringing inference closer to users in the Asia Pacific region. [AI News]
Alibaba Cloud has struggled to deploy servers fast enough to meet surging AI demand, leading to GPU rationing for its most comprehensive service users. [The Register]
Amazon is investing US$50B in AI and high-performance computing systems to enhance US government agencies' technological capabilities across various critical missions. [TechRadar]
AWS reportedly operated over 900 data centres across 50 countries in 2023, significantly more than previous estimates and potentially growing due to AI demand. [TechRadar]
Big Tech companies are rapidly pursuing data centre projects in rural Michigan, sparking local resistance over concerns about environmental impact, water usage, and community character. [MLive]
Brookfield Asset Management is launching a new AI infrastructure investment fund with Nvidia and KIA as investors, focusing on physical assets that support AI development. [Brookfield]
Google's AI infrastructure boss emphasised the need to double serving capacity every six months to meet growing demand for AI services. [CNBC]
Meta is seeking federal approval to trade electricity, aiming to accelerate power plant construction for its data centre energy needs. [TechCrunch]
Microsoft abandoned its data centre rezoning plan in Caledonia, Wisconsin, after facing significant local opposition, while successfully establishing a similar facility in nearby Mount Pleasant. [CNBC]
Alphabet is exploring selling its Tensor AI chips to Meta and other companies, potentially challenging Nvidia's market dominance. [Investor's Business Daily]
Nvidia dismissed potential Meta adoption of Google's TPUs, asserting its technological superiority and continued relevance in the AI accelerator market. [The Register]
The US Commerce Department approved Nvidia's sale of up to 70,000 AI chips to UAE and Saudi Arabian companies, marking a strategic diplomatic and technological move. [The Wall Street Journal]
Tachyum's Prodigy Universal Processor promises to reduce OpenAI's proposed US$3T, 250 GW data centre to a US$27 billion, 540 MW solution by 2028. [Business Wire]
Anthropic signed a US$50B data centre partnership with Fluidstack, aiming to build custom facilities in Texas and New York by 2026. [TechCrunch]
Meta may lose Yann LeCun, its renowned AI chief scientist, who plans to leave and start a world models-focused startup. [TechCrunch]
J.P. Morgan warned that the AI industry needs US$650B in annual revenue to deliver a 10% return on investments through 2030. [Tom's Hardware]
OpenAI released GPT-5.1, introducing two new models with enhanced conversational abilities and customizable conversation styles for ChatGPT users. [Gizmodo]
OpenAI is projecting massive annual losses through 2028, betting on explosive growth and infrastructure investment to reach US$200B in revenue by 2030. [Fortune]
Apple is finalising a US$1B annual deal with Google to enhance Siri using a powerful AI model. [Bloomberg]
Google is exploring a potential US$350B investment in Anthropic, expanding its existing stake and cloud computing partnership with the AI startup. [Sherwood]
Important Note:
The information in this article should not be considered an offer or solicitation to deal in ICM HPQC Fund (Registration number T22VC0112B-SF003) (the “Sub-fund”). The information is provided on a general basis for informational purposes only and is not to be relied upon as investment, legal, tax, or other advice. It does not take into account the investment objectives, financial situation, or particular needs of any specific investor. The information presented has been obtained from sources believed to be reliable, but no representation or warranty is given or may be implied that it is accurate or complete. The Investment Manager reserves the right to amend the information contained herein at any time, without notice. Investments in the Sub-fund are subject to investment risks, including the possible loss of the principal amount invested. The value of investments and the income derived therefrom may fall or rise. Past performance is not indicative of future performance. Investors should seek relevant professional advice before making any investment decision. This document is intended solely for institutional investors and accredited investors as defined under the Securities and Futures Act (Cap. 289) of Singapore. This document has not been reviewed by the Monetary Authority of Singapore.
ICM HPQC Fund is a registered Sub-fund of the ICMGF VCC (the VCC), a variable capital company incorporated in the Republic of Singapore. The assets and liabilities of ICM HPQC Fund are segregated from other Sub-funds of the VCC, in accordance with Section 29 of the VCC Act.
Get to know us better:
Comments