Arm CEO Rene Haas on AI: Nvidia Lessons, Intel’s Decline and the US-China Chip War

Summary

• The episode’s Arm setup was unusually strong. Its September IPO valued it above $54 billion, making it the largest public offering in over two years; the opening framing said the valuation had tripled. Later, the hosts put its market cap at $150 billion after SoftBank’s $32 billion take-private and failed sale attempt.
• Arm’s AI leverage is its CPU/IP layer rather than manufacturing. It increasingly supplies the microprocessor connecting accelerators such as Nvidia’s, Google’s and Cerebras’s; Nvidia’s Grace Blackwell uses “72 Arm CPUs.” Haas hinted Arm may go “a little bit further” than it does today, without confirming it will make chips or compete directly with Nvidia.
• Haas expects AI compute to split into training, dedicated inference, and a middle tier of smaller models that both learn and infer. Giant models could teach smaller, roughly 20-billion-parameter mixture-of-experts models, while endpoint inference cannot depend on a GPU “that runs at a kilowatt of power.”
• Haas said physical AI is already bigger than data centers today and could be huge by unit volume because each robot may contain tens or hundreds of chips. Today’s systems largely repurpose automotive silicon; future systems may need chips specific to actuators, joints and on-device learning.
• Haas’s lesson from Jensen Huang is Nvidia’s willingness to pivot quickly. He recalled Huang moving 2,000 of Nvidia’s roughly 6,000 employees from an Intel-linked chipset program into Arm-based SoCs: “What was intended to be a roadmap review turned into, ‘We’re changing the strategy.’”
• Intel’s decline illustrates how semiconductor mistakes compound across decade-long cycles. Missing mobile and underinvesting in EUV let TSMC attract Apple, Nvidia and AMD, improve through their volume, and widen the gap: “Once you fall behind in chips, it’s very, very difficult to catch up.”
• Rebuilding US semiconductor capacity requires industrial policy, corporate capital and manufacturing culture—not merely fab construction. Haas argued America has lost the “muscle memory” for 24/7 operational excellence and must restore manufacturing’s prestige through universities, corporations and financing sustained for years.
• Broad export licensing could create the rival technology ecosystem it is meant to contain. Haas warned that capable countries denied computing architectures “will find a way,” producing “two parallel universes” and putting the Western ecosystem at risk of losing global preference. He said China’s current software ecosystem largely follows the global one, including Android-derived mobile software and ADAS stacks, and argued for keeping that ecosystem open.
• Arm remains globally distributed and still needs more engineers. Half its employees are in the UK, with 2,000 in Bangalore and probably over 1,000 in the United States; Haas said AI has reduced finance and legal hiring but not engineering, and argued for more STEM investment.

Deep dive

1. Arm’s position—and Nvidia’s pivot

• The opening setup said Arm does not manufacture tangible products. Its September IPO valued it above $54 billion, the largest public offering in over two years, and the opening framing said its valuation had tripled. Later, the hosts put its market cap at $150 billion after SoftBank took it private for $32 billion, failed to find a bidder, and took it public again.

• Arm designs the processor/IP while others build the chips—mostly at TSMC, with some production at Samsung and even Intel. Haas described Arm as increasingly the CPU link between hardware and software and the processor connected to accelerators.

• Haas’s enduring lesson from Jensen Huang is the combination of “vision, speed, fearlessness, taking risks” and an ability to pivot extremely fast. When Nvidia was around $4 billion in sales, it was evaluating different ways to grow.

• The clearest example was an offsite where Nvidia abruptly changed strategy, abolished a product line and reassigned 2,000 of its approximately 6,000 employees. The effort involved mobile chipsets connecting to an Intel processor; competing with Intel’s integrated PC architecture was punishing. Haas said that episode helped prompt Nvidia’s major pivot toward SoCs and Arm-based architecture.

• Nvidia’s AI position began with workload fit, not a purpose-built AI chip: AlexNet training ran on a gaming GPU because training is massively parallel. Every AI workload still needs a CPU to run the computer and help the accelerator; Nvidia’s Grace Blackwell combines the Blackwell architecture with “72 Arm CPUs.”

2. Arm can serve every branch of a fragmenting AI market

• Haas said Arm can provide a standard solution or intellectual property for a custom chip, connecting to accelerators from Nvidia, Google, Cerebras and others. He did not confirm that Arm will itself make chips: when the host pressed that conclusion, he said he was not going to say that today, while noting that it was possible and that Arm was considering going “a little bit further than we do today.”

• Haas expects more than a training-versus-inference split. A giant model could teach smaller, roughly 20-billion-parameter mixture-of-experts models that combine inference, reinforcement learning and training—the “professor teaching a student who can also be a student-teacher.”

• At the edge, energy becomes decisive. Headsets, wearables and other endpoints cannot run a GPU at one kilowatt of power. Haas sees Arm as unusually positioned for these energy-efficient workloads.

• Haas said physical AI is already bigger than data centers today and could become enormous by unit count. Robots may contain tens or hundreds of chips. Current systems largely use repurposed automotive chips with functional-safety compliance for ADAS; future physical-AI systems may need chips specific to actuators or smaller parts of a joint, as well as chips that can learn.

3. Intel’s misses became TSMC’s compounding advantage

• Haas framed semiconductors as a business where long product, fab and ecosystem cycles punish even a few missed turns. Intel missed mobile and failed to invest in EUV at the rate TSMC did roughly a decade ago.

• The resulting flywheel is difficult to reverse: Apple, Nvidia and AMD manufacture at TSMC; their leading-edge work improves TSMC’s fabs. The host summarized the consequence by saying Intel and Samsung receive fewer opportunities. “Once you fall behind in chips, it’s very, very difficult to catch up.”

• Haas endorsed government support beyond an Intel stake, including upstream capabilities around ASML-class equipment and related infrastructure. He said access to rare-earth minerals is global; the bottleneck is refining the materials and building the factories, a decades-long investment.

4. American fabs need institutional patience and operational muscle

• Haas said he was impressed by China’s engineering-led industrial policy because it can last beyond an election cycle. His US prescription combined universities, corporations, private equity and other financing around initiatives whose capital requirements and timelines are too large for any one constituency.

• TSMC’s model is a “24/7 operation”: technicians and engineers must respond immediately when a line fails or a customer has a problem. America once possessed that manufacturing discipline but has lost both its “muscle memory” and the cultural prestige attached to such careers.

• Pressed for a practical solution, Haas pointed to universities rebuilding microelectronics and chip-design programs, including Carnegie Mellon. Manufacturing operations excellence should likewise become a formal discipline capable of rebuilding the workforce pipeline.

5. Export controls risk splitting the global compute ecosystem

• The host noted that an export-controlled advanced semiconductor sale requires a Commerce Department license and interagency approval. The process can take months, and some applications have remained pending for two years, by which time the chip is obsolete. The host warned against treating GPUs “like plutonium” and licensing advanced-semiconductor sales worldwide.

• Haas said the West’s compute leadership reflects both chip innovation and a global software ecosystem. The ecosystem works best when it is flat and unconstrained by restrictions on whom companies can sell to or how ecosystems can develop.

• If supply of a computing architecture is cut off, countries with sufficient people, technology or innovation “will find a way around the problem.” The danger is “two parallel universes,” with the alternative ecosystem potentially becoming the ecosystem of choice. Haas said that if licenses can be expedited, semiconductors work best as a global ecosystem where “may the best company win.”

• Haas also said China’s software ecosystem currently follows the global one: Chinese mobile phones use a version of Android and its app ecosystem, while autonomous vehicles leverage the ADAS stack that was created by Arm, and then Qualcomm and Nvidia. He said keeping the global ecosystem open was desirable.

• On China more broadly, Haas was an explicit optimist about collaboration. Based on conversations there, he believes Chinese officials view AI guardrails and policies as ways to maintain safety checks. He would not equate the situation with a nuclear arms race, but said there is a similar need for countries with the relevant capabilities to sit at the same table.

6. Arm’s origins and global talent model

• Arm began in a Cambridge barn as part of an Apple–VLSI Technology joint venture for the Apple Newton, which needed a low-cost, battery-powered processor. Haas recalled that the original chip “wasn’t so good,” but its design was good enough for the team to build a business around it.

• Haas is Arm’s fourth CEO and its first who is not from the UK. In the three and a half years since taking over, he has tried to preserve Cambridge’s scientific and technical strength while adding more Silicon Valley aggressiveness, speed and willingness to move quickly.

• Half of Arm’s employees are in the UK, with 2,000 in Bangalore, probably over 1,000 in the United States and others across Europe. Haas said the company goes where the engineering talent is.

• AI has reduced Arm’s hiring needs in finance and legal, but not engineering. Haas said AI development, creation and science remain difficult problems, so Arm needs more engineers and broader investment in STEM, electrical engineering and chip design.