Computing power subsidies are handing the future of general intelligence over to cloud giants.

Why Anderson’s Joke Landed: Because “Subsidy Moats” Are Real

The joke Marc Andreessen made about the AGI pricing gradient spread because it hit a real problem: compute subsidies have become a moat. He quoted a price of “negative $9 million” for “11 elite companies,” but in reality he’s raising an uncomfortable question—AGI development isn’t moving toward openness; instead, it’s concentrating among players who can get the cheapest compute.

I tried to cross-check the list of beneficiaries of the CHIPS Act with the cloud giants’ partners to verify who these “11” actually are, but I couldn’t find it. That number is more like a rhetorical device. But that precisely illustrates the point: this tweet, in a provocative way, forces investors to think—are policies quietly defining the winners of AGI?

The spread isn’t actually that wide: Grok summarized it, a small number of VCs discussed it, but mainstream AI voices like Karpathy and LeCun didn’t take the bait, so the impact is currently limited to the venture capital circle. Some supporting evidence can be considered: Arm’s AGI CPU partners are roughly about 9-10 companies, and CHIPS funding is still flowing to Intel and TSMC. But from April 2026 onward, no new policy disclosures directly pointing to AGI advantages have been issued. The timing Anderson chose lands right in this ambiguous zone—something that can be imagined but cannot be verified.

• Startups are being squeezed by costs: The idea that “external players’ prices equal infinity” isn’t entirely a joke. Compute is too expensive; small and medium teams can only rely on giants like Microsoft and Google, which have received subsidies.
• Talent flow is more consequential than policy wording: Instead of fixating on whether the “11 companies” list is correct, focus on how engineering and research talent is accelerating its flow into the “deep capital pool,” a structural risk.
• Open source and decentralization are even more urgent: If subsidies have already distorted the landscape of the AGI race, then the open-source strategy promoted by Meta and others will compress the premium pricing cycle even faster.

How Different Camps See This (And What It Means for Pricing)

Divergences reflect potential mispricing space:

What’s interesting is that the collective silence from mainstream experts actually amplifies the effect of this tweet: it shifts investors’ attention from “can we build AGI?” to “who controls the most restrictive subsidized compute?” And the “11 companies” that weren’t named become a shorthand for big-tech leadership.

But this argument also has loopholes: without a verifiable company list and new policy evidence, this satire relies more on inference, making it easy to overestimate barriers and underestimate the buffer provided by pathways like federated learning.

Key Takeaways:

• The moat is indeed narrowing, but the market’s pricing of decentralization and edge computing is clearly lagging.
• Enterprises that can directly leverage big-tech infrastructure are in the best position; other players should place bets on open-source and decentralized alternatives as soon as possible.

Importance: Moderate
Category: Industry Trends, Market Impact, AI Policy

Assessment: For this topic, the overall reaction of funds in the secondary market is relatively slow. The biggest beneficiaries are the buyer-side companies that can immediately connect with big cloud giants’ infrastructure, as well as the Builders that do hardware and edge distribution; passive holders and short-term Traders chasing “AGI concepts” are at a disadvantage. Funds should prioritize allocating incremental positions to open-source and decentralized infrastructure, rather than to a “pure AGI narrative.”