Competition, collaboration or collusion: implications for the stack
Updated: 14 hours ago
By Oliver Campbell │ ICM HPQC Fund │ September 2026
“Written by humans, please don’t blame the robots for our typos”
Twenty years ago the financials of the compute supply chain looked very different to those of today. The end-customer or branded original equipment manufacturer (OEM) typically held the strongest financial position in the chain and squeezed suppliers as hard as possible: running negative working capital, demanding double-digit annual price (ASP) declines that forced suppliers into low profitability, and offering them high asset turns as the only lever left to claw returns back up.
The power of sitting atop a supply chain
Today the picture is different but only in select areas. Supply chain commentary from branded fabless OEMs is now far more about sustainable collaboration than cut-throat, zero-sum competition. In turn, some suppliers have shifted up the value chain and are now sitting on large cash balances, running higher margins and higher structural returns than in the past. But this tends to be true only of the suppliers sitting on genuine chokepoints: extremeultraviolet (EUV) lithography, leading-edge foundry capacity, high bandwidth memory (HBM), data centre power supply, among others. While the higher-volume, lower-mix businesses in more commoditised parts of the chain are still fighting for margin, cashflow and returns.
Improving fortunes in selected areas of the stack
There are three implications for HPQC's current and future holdings, and for what they must seek out deliberately rather than assume will happen to them. This might get a bit dry, but it's important, so hang in there.
1. Find the chokepoint, and invest heavily in defending it
The single biggest determinant of where a company sits on the margin spectrum in this supply chain is not how large its addressable market (TAM) is, but how narrow the physical bottleneck is that it controls. EUV lithography has essentially one supplier - ASML. Leading-edge foundry capacity has, in practice, one and a half. HBM has three, and even that is a recent widening. Data centre power, interconnect, grid capacity, and cooling are emerging as the newest and least-priced chokepoint of the artificial intelligence (AI) build-out.
For HPQC, this means investment decisions should start from the question “what is the physical sub-constraint this company owns or is closest to owning,” not “how big is the TAM.” A company can be early into a large market and still earn commodity-level returns if the thing it makes can be substituted or second-sourced. The companies worth backing heavily are the ones where substitution is genuinely hard because of physics, yield, intellectual property (IP), or multi-year qualification cycles and where that difficulty is getting harder, not easier, as the underlying technology advances (smaller nodes, denser memory stacks, higher power density all tend to narrow the chokepoint further rather than widen it).
Two portfolio examples make this concrete. Diraq deliberately doesn't own a fab. Its chokepoint is the twenty years of IP behind getting silicon spin qubits to work reliably on standard CMOS processes, evidenced by its 99%+ two qubit gate fidelities and a growing patent estate. The substitution barrier isn't manufacturing access, which it rents from imec's 300mm line; it's the accumulated know-how in making CMOS-compatible qubits behave, which is very hard to replicate on any comparable timeline.
Meeting Diraq in UNSW, Sydney, August 2026
2. Strive to avoid commoditisation
The flip side of point one is knowing when a portfolio company is drifting into a segment where several credible suppliers can do the same thing to the same specification. This is the fate that befell DRAM (memory) and TFT-LCD (display) makers two decades ago, and it is the fate that continues to await server assembly, generic packaging, and commodity power-supply units today. The warning signs are familiar: falling ASPs that aren't offset by falling costs, customers qualifying a second source, and a widening gap between the company's margins and those of the chokepoint suppliers just upstream of it.
HPQC's diligence and portfolio-monitoring process should treat this as a recurring question, not a one-time investment check: a company's chokepoint position can erode over a holding period even if nothing about the company itself has changed, simply because the rest of the supply chain caught up.
Hon Hai Dupont declining returns analysis
Two portfolio companies illustrate the two ways to sidestep this. SpiNNcloud avoids commoditisation by refusing to compete on the terms of the market it's adjacent to: rather than fighting Nvidia and AMD on FLOPS-per-dollar in a GPU (graphical processing unit) race it can't win, its SpiNNaker2 architecture competes on an entirely different axis: event-driven, energy-proportional computation for workloads that GPUs are structurally inefficient at.
Q-CTRL avoids a narrower version of the same trap: because its control and error-suppression software is hardware-agnostic and already integrated across IBM's, Rigetti's and other providers' stacks, its commercial fate isn't tied to which qubit modality (superconducting, trapped-ion, silicon spin, etc) ultimately wins the underlying hardware race. This represents a structurally different risk profile from a single-modality hardware bet, and one worth weighing explicitly when sizing positions across the quantum sub-portfolio.
3. Seek healthy supply chain partners for capital, for customers, and for exit
Financially strong upstream and downstream partners are, on the whole, good news for HPQC's holdings, not a competitive threat. A cash-generative supply chain partner is more likely to invest in, co-develop with or acquire a chokepoint-holding portfolio company than build the capability in-house, because partnering preserves its own margin structure rather than diluting it (more on this below).
The practical implication is that HPQC should actively map which of its portfolio companies' upstream and downstream partners are strengthening financially and treat that strengthening as a signal to deepen the relationship through co-investment, supply agreements, or board-level engagement, well before an exit process begins, since these partners are increasingly the natural acquirers.
The portfolio already has three ongoing versions of this. Diraq rents its manufacturing chokepoint from imec rather than building one, which only works because imec is itself a healthy, well-capitalised R&D foundry with every incentive to keep proving out next-generation processes with partners like Diraq rather than compete with them.
Salience Labs has taken this further into a formal manufacturing partnership with Tower Semiconductor to move its photonic switches from development into volume production, precisely the welcoming, well-capitalised partner the thesis calls for, rather than a foundry relationship Salience would have to fight for capacity on.
Mixx Technologies is following the same logic geographically: its build-out of manufacturing and operations in Taiwan is a deliberate move to sit physically closer to the healthiest, highest-margin part of the supply chain (TSMC, the packaging houses, the assembly ecosystem) rather than trying to replicate that infrastructure from the US or India.
HPQC meeting Mixx Technologies at Semicon Taiwan, September 2026
It's also worth noting that, for several of these companies, the “healthy partner” relationship shows up directly on the cap table, not just in a commercial contract, which is arguably a stronger commitment. Diraq's investor base includes imec.xpand, imec's own corporate venture arm, alongside In-Q-Tel: the manufacturing partner has put capital in as well as capacity.
Q-CTRL's Series B syndicate reads like a map of exactly the kind of supply-chain-adjacent strategics the thesis is describing: Airbus Ventures, Lockheed Martin Ventures, NTT Finance and Salesforce Ventures have all invested, none of them a competitor, all of them customers or ecosystem partners with a direct interest in Q-CTRL's hardware-agnostic software making the rest of their own stack work better.
Mixx's register includes TDK Ventures; TDK being a components and materials supplier into exactly the AI-infrastructure build-out Mixx sells into, alongside an earlier, more direct signal: Kaynes Technologies, an electronics manufacturing services company, bought a 13% equity stake in Mixx in January 2024, a genuine assembly-partner-as-investor relationship rather than a pure financial one.
Won't stronger supply chain companies compete with HPQC's holdings?
The intuitive worry is that a cash-rich, structurally strengthening supplier, flush with the kind of returns TSMC or SK Hynix are now generating, will eventually decide to build the capability itself rather than buy it from a smaller chokepoint-holding company, competing HPQC's portfolio out of existence from a position of overwhelming capital advantage.
In practice this rarely makes sense for the stronger partner, and Nvidia's own capital allocation is the clearest illustration of why. Nvidia generated roughly $49 billion of free cash flow in a single quarter in FY2027i,has committed to returning roughly half of free cash flow to shareholders, repurchased ~$40 billionii of stock in FY2026 alone, and as of mid-2026 had close to $100 billion in buyback authorisation still outstandingiii — all while continuing to grow R&D spend to record levels each quarter.
A company with that cash-generation profile has every incentive to keep deploying capital into the areas that already earn it ~75% marginsiv (its own chip design and platform business) while keeping the rest of the stack - packaging, HBM, power, assembly - as an ecosystem of partners it invests in, co-develops with, and depends on, rather than bringing those dilutive, lower-return capabilities in-house. Vertically integrating into a 2–3% margin assembly business, or a capital-intensive fab business someone else already does well, would be value-destructive for a company earning Nvidia's returns. The maths simply doesn't work in the acquirer's favour, even when the short-termism of the stock market prefers capitalised M&A over expensed R&D.
The one exception worth flagging is Nvidia’s own 2020 Mellanox acquisition, now a $31 billion+ annual businessv; proof the “partner not compete” logic holds for commodity layers like assembly, but not for interconnect, where the chokepoint is real enough that Nvidia chooses to build/buy rather than partner.
The same logic applies further down the chain: TSMC has no reason to compete with the packaging or substrate specialists it depends on, and Hon Hai or other ODM and EMS companies have every reason to deepen their AI-server partnerships rather than try to out-invest existing suppliers. Strength begets partnership between adjacent tiers far more often than it begets vertical competition, because each tier's return profile depends on staying focused on the part of the stack where it already has pricing power.
What this means for HPQC's portfolio companies
Put together, the three implications above aren't independent, instead they compound. Companies genuinely sitting on chokepoints are, almost by definition, operating in the part of the stack where margins, growth and cashflow durability are all improving simultaneously, not trading off against each other. The strengthening of the broader supply chain around them is a tailwind, not a threat, because stronger neighbours have both the balance sheet and the strategic incentive to partner, invest in, or eventually acquire chokepoint-holding companies rather than compete with them.
And because that strengthening shows up directly in the acquirers' own cashflow, as it has for TSMC, for memory, and now visibly for Nvidia's balance sheet, it also shows up in what those acquirers can and will pay for the chokepoint assets they don't yet own. Higher cashflows upstream and downstream of HPQC's holdings could, over a holding period, translate into higher achievable valuations at exit for the fund, provided the underlying chokepoint position is defended rather than allowed to commoditise in the meantime.
Oliver Campbell
Investment Principal of the ICM HPQC Fund
MAS Licensed Representative, ICM Global Funds Pte Ltd
September 2026
Sources:
[i] Nvidia 10Q, https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-first-quarter-fiscal-2027, levered free cashflow calculated as adjusted operating cashflow minus capital investments
[iv] ibid
Important Note:
The information in this article should not be considered an offer or solicitation to deal in the ICM HPQC Fund (Registration number T22VC0112B SF003) or ICM HPQC Fund 2 (Registration number T22VC0112B SF007) (the “Sub-funds”). 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. Investors should seek relevant professional advice before making any investment decision. 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-funds are subject to investment risks, including the possible loss of the principal amount invested. All forms of investments carry risks, including the risk of losing all of the invested amount. Investors should read the prospectus before deciding whether to acquire the units in the Sub-funds. The value of investments and the income derived therefrom may fall or rise. Past performance is not indicative of future performance. This document is intended solely for institutional investors and accredited investors as defined under the Securities and Futures Act 2001 of Singapore. The whole or any part of this work may not be reproduced, copied or transmitted or any of its contents disclosed to third parties without ICM Global Fund’s express written consent. This advertisement or publication has not been reviewed by the Monetary Authority of Singapore.
ICM HPQC Fund and ICM HPQC Fund 2 are registered Sub-Funds of the ICMGF VCC (the VCC), a variable capital company incorporated in the Republic of Singapore. The assets and liabilities of ICM HPQC Fund and ICM HPQC Fund 2 are segregated from other Sub-Funds of the VCC, in accordance with Section 29 of the VCC Act.
Get to know us better:
Comments