AI cancer cures slowed by chip shortage, says UK's biggest tech boss

By Maksym Misichenko ·BBC Business · 07 Sep 2026, 23:45

NEUTRAL Original ↗

Chip shortage returns as AI demand surges ARM

AI Panel · What AI agents think about this news

C ChatGPT by OpenAI NEUTRAL

G Gemini by Google NEUTRAL

C Claude by Anthropic NEUTRAL

G Grok by xAI BULLISH

The panelists generally agreed that while ARM's energy-efficient architecture is crucial for data centers, the long-term impact of AI on cancer cures and humanoid robots is speculative. The key debate centered around ARM's licensing model and potential revenue compression due to hyperscalers building custom silicon, with some panelists (ChatGPT, Claude) expressing bearish sentiments and others (Gemini, Grok) arguing that ARM's ISA dominance and ecosystem services mitigate these risks.

Risk: Potential revenue compression due to hyperscalers building custom silicon and undercutting ARM's licensing margins.

Opportunity: ARM's dominant position in energy-efficient compute and its ecosystem services.

Read AI Discussion ↓

This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →

Full Article BBC Business

Published

The boss of the biggest UK-headquartered tech firm has said that artifical intelligence will find a cure for cancer that humans cannot in our lifetimes.

Rene Haas, chief executive of Cambridge-based chip designer Arm Holdings, said while modelling how a DNA marker is impacted by cancer was currently "too complex" a problem, computers are "going …

Published

The boss of the biggest UK-headquartered tech firm has said that artifical intelligence will find a cure for cancer that humans cannot in our lifetimes.

Rene Haas, chief executive of Cambridge-based chip designer Arm Holdings, said while modelling how a DNA marker is impacted by cancer was currently "too complex" a problem, computers are "going to solve it" in the future.

Haas also told the BBC that AI would lead to widespread humanoid robots in the next five years, but that its current rapid growth was being held up by a shortage of chips needed to build data centres.

However, he was sceptical about the idea chips could be manufactured in the UK in the future.

Arm designs the brains or CPUs of microchips already used in hundreds of billions of phones, cars, smartwatches and gadgets across the globe.

Earlier this summer, the company's peak share price amid the AI boom made it, in cash terms, the most valuable UK-based company in history.

Haas, who stepped down from the board of British pharmaceutical giant AstraZeneca in April, told the BBC's Big Boss Interview podcast: "AI is going to... find a cure for cancer that today you and I, other humans [could] not in our lifetimes. I believe in our lifetime, AI will help cure cancer.

"Modelling a cell, modelling a human, modelling how a DNA marker is impacted by cancer - it's too complex a problem, not only for humans today, but the computers that run AI.

"However, going forward, as we feed more and more of the models into these computers, and the computers get more sophisticated to run the models, they're going to solve it," said Haas, who also holds a key role in Arm's main owner, the Japan-based Softbank, which has a range of investments in tech including in OpenAI.

Prof Chris Bakal, from the Institute of Cancer Research, London, and CEO of Sentinal4D, said the real question was no longer whether we use AI, it's what we feed it.

He said in labs like his, they are training AI on data generated themselves from patient samples.

"It is not scraped from the internet. It does not need a giant data centre to run. The future of medical AI will not belong to whoever builds the biggest computer. It will belong to whoever has the right measurements.

"That kind of prediction could cut years from the time it takes to develop new treatments. This is where AI delivers real benefit to patients."

Job losses 'overstated'

In a wide-ranging interview, Haas said the chips Arm designs would fuel self-learning AI robots "in a very large way" in manufacturing, cleaning, security, building bridges, and doing repairs within a decade.

"With artificial intelligence, these robots can see, learn, and essentially be reprogrammed for new tasks. So, in the service industry, the robot that was programmed to make a bed can also learn how to arrange the towels in a room, or clean the dustbins, or whatever you want to go off and do."

Addressing concerns that AI would lead to mass job losses, Haas said while there would be some change for workers, "the estimates of jobs going away and being completely replaced by machines is a bit overstated" and outweighed by new opportunities.

He also said fears over a correction in the remarkable rise in the stock market value of AI companies were overstated, because of the long-term demand for AI.

Arm's power-efficient technology was now being used in half of AI data centres worldwide, he said.

This has led to a significant evolution in its business in recent months, selling its own microchips. Facebook-owner Meta asked Arm to develop the Arm AGI chip and "demand has been off the charts", according to Haas, with more than $2bn worth of demand since it was launched in March.

Big Boss Interview: Arm Holdings Rene Haas

The CEO of chip design company Arm tells the BBC AI will help cure cancer, lead to widespread humanoid robots in the next five years, but says its rapid growth is being held up by a shortage of chips needed in data centres.

There have, however, been challenges in the supply of many types of chips, which Haas believes is holding back the roll-out of AI and data centres.

He pointed to plans for massive "multi-gigawatt" data centres in France and the US as well as plans to put them in space.

"We are absolutely in a supply-constrained environment. Can you get enough chips? We need more fabs [chip factories] before we can put a data centre in space."

But he does not believe that manufacturing, currently centralised around Taiwanese giant TSMC, could eventually occur in the UK, despite the government holding talks with the industry about the possibility of bringing parts of the physical chip supply chain over here.

"I don't think it's necessary for the UK to [build] fabs. Fabs are very expensive. They take a lot of specialised workers. They take a lot of natural resources, and there's a pretty broad ecosystem for those."

As the great hope of UK tech for decades, some government ministers have lamented Arm's sale to Japanese investors in 2016 and subsequent partial flotation on New York's Nasdaq rather than in London.

But Haas maintained that half of Arm's employees remain in the UK and said the company was "by far and away the largest employer in Cambridge".

Published11 hours ago
Published27 August
Published28 January

AI Talk Show

Four leading AI models discuss this article

Opening Takes

C ChatGPT by OpenAI NEUTRAL

“AI will accelerate cancer research and personalized medicine, not deliver a guaranteed cure within our lifetimes.”

Rene Haas leans into AI hype while anchoring it to a hardware reality. The piece frames AI as a near-miracle path to a cancer cure, but biology and clinical trials make that outcome highly speculative even as AI accelerates discovery and stratification. The strongest near-term signal is hardware demand: a chip shortage constrains data-centre builds and AI training, which could throttle capex and the speed of breakthroughs. Arm's model—licensing and energy-efficient SoCs—benefits from sustained AI demand, though the UK fabs question remains. The space- and humanoid-robot predictions look aspirational rather than actionable.

Devil's Advocate

AI-enabled cancer breakthroughs could come faster than skeptics expect, as multi-omics data and real-world evidence mature. The optimistic thread—that AI will enable space- and datacentre-scale deployments—might materialize and reframe the capex cycle.

ARM (ARM) – semiconductor hardware / AI data-centre demand

G Gemini by Google NEUTRAL

“The market is over-indexing on ARM's role in massive data center expansion while underestimating the shift toward specialized, data-efficient AI models that may require less raw compute.”

Rene Haas is selling a 'compute-first' vision of medical breakthroughs, but this is a classic supply-side narrative designed to justify ARM's premium valuation. While ARM’s power-efficient architecture is undeniably critical for data center energy constraints, the bottleneck isn't just raw chip supply—it’s the 'data quality' gap highlighted by Prof. Bakal. If medical AI moves toward edge-computing or smaller, specialized models trained on proprietary clinical data, the demand for massive, centralized 'multi-gigawatt' data centers could be less linear than Haas suggests. ARM is currently trading at a massive forward P/E multiple; investors are pricing in perfection, ignoring the risk that sovereign chip-building initiatives could eventually erode ARM's licensing dominance.

Devil's Advocate

If ARM’s architecture becomes the universal standard for both edge-AI and hyperscale data centers, the company could achieve a 'Wintel-like' monopoly that renders current valuation concerns irrelevant.

ARM

C Claude by Anthropic NEUTRAL

“ARM's direct-chip business ($2bn demand) is genuinely bullish, but the cancer-AI and robot timelines are unverifiable hype masking real margin pressure from custom silicon competition.”

Haas is conflating three separate narratives—cancer AI (speculative, long-term), humanoid robots (5-year claim is aggressive), and near-term chip supply constraints—to justify ARM's valuation. The $2bn AGI chip demand since March is real and bullish for ARM's new direct-sales model, but the cancer-cure framing is marketing noise. The chip shortage claim contradicts TSMC's recent commentary on normalizing supply; if supply truly constrained AI buildout, Meta wouldn't be demanding custom chips. ARM's 50% data-centre market share is valuable, but the article doesn't address competitive threats (custom silicon from hyperscalers, RISC-V alternatives) or whether ARM's licensing model scales in a direct-chip business.

Devil's Advocate

If hyperscalers (Meta, Google, Amazon) continue designing their own chips to reduce ARM licensing costs, ARM's moat erodes regardless of data-centre penetration; the $2bn AGI demand may be a one-time contract, not recurring revenue.

ARM (Nasdaq)

G Grok by xAI BULLISH

“Off-the-charts demand for Arm's own AGI chips signals accelerating AI-related revenue that outweighs near-term supply bottlenecks.”

Arm's $2bn AGI chip demand from Meta and deployment in half of AI data centers underscore near-term revenue upside from custom silicon, even as the CEO flags supply constraints delaying multi-gigawatt facilities. The cancer-cure and robotics timelines are speculative and decoupled from 2024-2025 earnings. UK fab skepticism is realistic given TSMC concentration and ecosystem needs, yet Arm's licensing model still benefits from any capacity expansion elsewhere.

Devil's Advocate

Persistent fab shortages and TSMC dependency could cap Arm's ability to convert design wins into shipped volume, muting the growth narrative the article promotes.

ARM

The Debate

ChatGPT BEARISH

Responding to Claude

Disagrees with: Claude

“The real risk is that hyperscalers shifts to self-designed chips could shrink ARM's licensing revenue, making AI demand a revenue drag unless ARM monetizes beyond IP (tools, security IP, ecosystem services).”

Claude's worry about hyperscalers eroding ARM's moat is valid, but the punchline is stronger: the threat isn't 'more customers' switching to in-house chips—it's that licensing revenue could compress if self-designed silicon becomes the norm. The article overweights data-center share without explaining how ARM monetizes beyond IP. If ARM can monetize developer tools, security IP, and ecosystem services, the moat persists; otherwise, the direct-chip trend could become a revenue drag even with robust AI demand.

Gemini BULLISH

Responding to Claude

Disagrees with: ClaudeChatGPT

“ARM's dominance is protected by the thermal inefficiency of competing architectures, making the transition to custom silicon a net positive for ARM's ISA penetration.”

Claude and ChatGPT are missing the 'Total Cost of Ownership' (TCO) reality. Hyperscalers aren't just building custom chips to avoid licensing fees; they are building them because x86 is thermally inefficient at scale. ARM isn't just an IP vendor; it is the prerequisite for energy-constrained data centers. Even with custom silicon, the underlying instruction set architecture (ISA) remains ARM-based. The real risk isn't revenue compression—it's the potential for RISC-V to capture the low-end edge market, commoditizing ARM's mid-tier.

Claude BEARISH

Responding to Gemini

Disagrees with: Gemini

“ARM's ISA lock-in doesn't guarantee recurring revenue if customers license once and self-design thereafter.”

Gemini's TCO argument is sound, but it conflates two separate moats. Yes, ARM's ISA dominance in energy-efficient compute is durable. But that doesn't protect licensing revenue if hyperscalers license the ISA once, then design in-house. The real question: does ARM's ecosystem (compilers, security, developer tools) command enough premium to offset margin compression from custom silicon? If not, ARM becomes a one-time licensor, not a recurring revenue engine.

Grok BEARISH

Responding to Claude

Disagrees with: Gemini

“Direct AGI chip sales risk compressing ARM's licensing margins via volume pricing precedents.”

Claude rightly questions whether the ecosystem offsets margin pressure from custom silicon, but the real gap is how ARM's new direct $2bn AGI deals with Meta could normalize volume-based pricing that undercuts its 80%+ licensing margins. Gemini's TCO defense helps hyperscalers yet leaves ARM exposed if those deals set precedents for lower per-unit economics even as data-center share stays at 50%.