Chamath’s claim is explicitly long‑dated (“over the next several years”) and about where most of the economic surplus in AI will ultimately accrue (infra/cloud and apps vs. model providers). As of late 2025, both the time horizon and the competitive dynamics are still in flux.

Evidence that hyperscaler/cloud infrastructure is capturing a lot of value:

• Microsoft reports very strong growth and profits in Intelligent Cloud: Azure revenue grew more than 30% year‑over‑year and helped push Microsoft Cloud revenue to tens of billions per quarter, with management explicitly attributing this to AI workloads.(news.microsoft.com)
• Amazon Web Services (AWS) is growing ~20% year‑over‑year on a >$130B annualized run rate, with AI and generative AI cited as key demand drivers; AWS contributes a disproportionate share of Amazon’s operating income and is backed by an AI‑driven backlog around $200B.(ir.aboutamazon.com)
• Alphabet’s Google Cloud and broader AI offerings have helped drive a ~70% stock rally in 2025 and pushed Alphabet’s market cap toward $4T, with analysts calling it a major AI beneficiary as cloud and AI tools become core profit drivers.(reuters.com)

Evidence that model providers themselves are also capturing enormous (though often unprofitable) value:

• OpenAI has raised tens of billions of dollars at valuations around the low‑hundreds of billions, with revenue already in the low‑teens of billions annually but very large projected operating losses. HSBC estimates OpenAI has signed cloud rental agreements worth hundreds of billions of dollars with Microsoft and Amazon and will need to raise over $200B by 2030 to fund compute and operating losses, underscoring both its scale and its capital intensity.(roic.ai)
• Anthropic has raised more than $13B, reaching a valuation around $183B with estimated 2025 annualized revenue around $5B and aggressive growth projections toward tens of billions in the late 2020s. It still expects to break even only around 2028, meaning much of the value is in anticipated future cash flows rather than current profits.(cnbc.com)

Evidence that hardware and models are not yet commoditized, contrary to the premise of the prediction:

• Nvidia has become the single biggest financial winner of the AI boom so far, surpassing a $5T market cap with data‑center revenue up roughly 9x in two years, and is widely described as the backbone of AI infrastructure with a strong software moat (CUDA). This is the opposite of a commoditized hardware supplier.(nasdaq.com)
• Hyperscalers themselves are pouring enormous capex into proprietary AI chips and data centers (e.g., AWS’s multi‑gigawatt expansion and custom Trainium chips, Google’s TPUs), which suggests a continuing struggle for differentiated, not yet commoditized, infrastructure.(earningsiq.co)

Netting this out as of November 30, 2025:

• Part of the prediction looks directionally plausible: cloud/infrastructure providers (Azure, AWS, Google Cloud) are clearly making substantial, real money from AI, with strong revenue and profit contributions.
• Model providers are also clearly capturing very large economic value via enormous valuations and rapidly growing revenue, even if many are still deeply loss‑making and heavily dependent on cloud partners. Whether, over the full multi‑year horizon, their investors will capture more or less value than the hyperscalers and application companies is still unresolved.
• The key assumption that models and specialized hardware will be commoditized is not yet borne out; Nvidia and a handful of top labs remain highly differentiated and richly valued.

Because (1) we are only about two years into a “several‑year” horizon, and (2) the ultimate distribution of profits between infra/cloud, applications, and model providers is still evolving with strong arguments on both sides, there is not yet enough evidence to say the prediction is clearly right or clearly wrong. It is therefore too early to call, so the outcome is best classified as inconclusive at this time.