James Anderson, the legendary tech investor who helped build Baillie Gifford into a technology powerhouse, is sounding an alarm that’s sending ripples through Silicon Valley? His warning about Nvidia’s planned $100 billion investment in OpenAI carries the weight of someone who’s seen this movie before�and remembers how it ended last time?
The Dotcom D�j� Vu
“I have to say the words ‘vendor financing’ do not carry nice reflections to somebody of my age,” Anderson told the Financial Times, drawing direct parallels to the dotcom-era practice where telecom equipment makers borrowed heavily to finance their customers’ internet build-outs? “It’s not quite like what many of the telecom suppliers were up to in 1999-2000 but it has certain rhymes to it?”
What’s particularly striking is Anderson’s dramatic shift in perspective? Just last year, he predicted Nvidia could reach “double-digit trillions” in market cap? Now, he’s trimming his position in the chipmaker while watching OpenAI’s valuation surge from $157 billion to $500 billion in less than a year? His main concern? The “circular structure” where Nvidia�OpenAI’s biggest chip supplier�is also becoming its biggest investor?
The Infrastructure Reality Check
Barclays analysts provide crucial context for Anderson’s concerns? Their research reveals that data center capital expenditure is growing at 30% annually into the next decade, with Nvidia chips accounting for 50-65% of AI data center costs? More alarmingly, these expensive chips have a useful life of just two years before becoming obsolete?
The US Department of Energy forecasts 2-3x growth in data center electricity demand by 2028, creating potential bottlenecks that could stall the AI revolution? Barclays calculates that a 20% cut in data center spending could lead to a 10-13% de-rating for the S&P 500�a sobering reminder of how interconnected this ecosystem has become?
The Stargate Gambit
OpenAI isn’t just thinking big�it’s thinking massive? The company has agreements covering approximately $1 trillion in capital spending, with five new Stargate data centers potentially costing $400 billion? Oracle alone is due $300 billion in cloud payments by 2030, while Nvidia deal analysts project $350-400 billion in new chip sales?
Sam Altman, OpenAI’s CEO, promises to lay out detailed financing plans later this year, but the scale raises fundamental questions? Can any company justify this level of infrastructure investment? And what happens if the anticipated AI productivity gains don’t materialize as quickly as expected?
The Productivity Paradox
This is where the rubber meets the road? Princeton professors Arvind Narayanan and Sayash Kapoor argue that “AI is a normal technology,” suggesting investors might be better off with broad market indices rather than betting on specific AI companies? Current productivity increase estimates range from 1?4% growth in a year to just 1% over five years�hardly the revolutionary transformation some are banking on?
Then there’s Jevons’ paradox: technological efficiency often leads to higher overall consumption rather than cost savings? As AI makes certain tasks cheaper, we might simply do more of them, maintaining or even increasing total spending?
The Performance Question
OpenAI’s recent GDPval benchmark offers some hope? GPT-5 was ranked better than or on par with human experts 40?6% of the time across 44 occupations�a significant jump from GPT-4o’s 13?7% score just 15 months ago? Dr? Aaron Chatterji, OpenAI’s chief economist, notes that “people in those jobs can now use the model to offload some of their work and do potentially higher value things?”
But here’s the catch: these benchmarks only test report submissions, not full job tasks? The real-world implementation might be more complex and less transformative than laboratory results suggest?
The Investor’s Dilemma
So where does this leave businesses and investors? Anderson’s caution contrasts sharply with the optimism driving current valuations? Barclays analysts summarize the situation perfectly: “We remain constructive on AI as an investment theme, but explore some ‘what-ifs’ for data center capex downside?”
The fundamental question isn’t whether AI will transform industries�it already is? The question is whether current valuations and investment levels reflect realistic timelines and returns? With hyperscaler capital expenditure at about 25% of sales (compared to over 40% during the dotcom bubble), there’s room for optimism? But with global network utilization at just 26% in 2022, there’s also room for massive overcapacity?
As Anderson puts it, those “sudden increases in valuation that people were willing to place on OpenAI, Anthropic and the like were disconcerting?” The scale and pace of the jumps “did bother me?” For someone who made his name betting big on technology’s future, that’s a warning worth hearing?