Arm's co-founder skips the bubble call and names the mechanism: circular financing

Hermann Hauser, who co-founded Acorn Computers in 1978, helped create Arm, and now backs European deep-tech startups through Amadeus Capital, told CNBC's The Tech Download podcast that AI is "a revolution that will create more value than probably any other technology revolution that we've ever seen" ^[1][2][3]. He also called it a "rollercoaster", and the specific risk he named was structural rather than sentimental: recently announced circular financing deals ^[4][5].

That is a narrower position than a bubble call, and a more usable one. On prices Hauser said only that some valuations have "clearly gotten ahead of themselves", a judgement anyone can take or leave ^[6]. Circular financing is a description of plumbing: a chipmaker takes a stake in an AI lab that then buys its chips ^[7]. The Bank for International Settlements flagged the pattern in June, warning that an AI bust could hit credit markets as hard as 2008 ^[8].

The two warnings do different work for anyone signing a multi-year compute commitment. A bubble call is about timing and says nothing about which counterparty breaks first. Circular financing is about whether a demand signal is independent of the party supplying the capital behind it. Where a vendor's equity funds a customer's order, the resulting revenue confirms that money moved, not that an unrelated buyer wanted the product at that price.

Hauser is not forecasting the 2008 comparison. He argues the largest players survive a reset, and on his account OpenAI and Anthropic hold significant capital reserves and should withstand turbulence even if expectations reset around them ^[9][10]. The write-up by Arjun Kharpal and Kai Nicol-Schwarz, published on 14 August, notes that he does not say who absorbs the reset instead, and that the circular deals are where it would land ^[11][12].

Set that beside Masayoshi Son, who told SoftBank shareholders in June that calling AI a bubble is an insult and has also called such commentary blasphemy ^[13]. Hauser is separating the technology question from the pricing question and answering them in opposite directions ^[14].

The hardware argument underneath is the part with prices already attached. Hauser thinks AI is forcing a rethink of computing architecture itself, on practical grounds: AI is expensive to run, chips are difficult to cool, memory is expensive, and bottlenecks run across the industry ^[15][16]. He pointed to in-memory computing and photonic computing, both of which target the energy spent moving data between processor and memory rather than the cost of the arithmetic ^[17]. In-memory computing moves the arithmetic into the memory instead of shuttling data to a processor and back ^[18]. "I never thought that we'd have a very fundamental change in the computer architecture as a result of AI," he said ^[19]. His precedent is Arm, which won on power efficiency while the industry optimised for raw speed ^[20].

Nvidia spent $6.5bn across photonics companies in three months to replace copper with light inside AI data centres ^[21], an average of roughly $2.2bn a month ^[22]. That is a bottleneck being bought rather than engineered around.

On Europe he is blunt in both directions: the companies have the innovation and skill to compete with the US and China, and they struggle to grow from small startup into a genuinely global competitor ^[23][24]. He warns that Europe depends on foreign suppliers for critical technologies, a range running from AI models to semiconductor design software ^[25].

Two things to watch. First, whether large compute agreements disclose the supplier's equity position in the buyer, because that is the fact Hauser's warning turns on ^[7]. Second, whether the photonics buying continues at anything like the pace Nvidia set in that three-month window ^[21], since the cost of moving data is the constraint both of his candidate architectures are built around ^[17].