Key Takeaways

  • Foundational intelligence will commoditize like the early internet, leaving value in the hands of consumers and operational businesses rather than raw model hosts.
  • Meta holds the best position to sustain profits because distribution channels like Muse pair directly with proven consumer product execution.
  • Standalone model makers face collapsing software margins unless they branch out into proprietary hardware or launch internal vertical corporations.
  • Large enterprises will gladly pay billions for brute-force compute: Martin Shkreli notes McDonald's could spend $10 billion on 10 million GPUs if the data yields a higher net present value for their next menu item.

The Cisco Trap and Free Intelligence

Silicon Valley treats raw model performance like a permanent moat. History says the opposite. Shkreli points out that the economics of generative software mirror the rollout of the commercial internet in the late 1990s. When underlying infrastructure matures, the rail providers rarely capture the bulk of the economic surplus.

“The only winners are going to be consumers and businesses,” Shkreli says. “They get this tax-free kind of internet. Cisco and some other guys will get a little bit out of it, but the main highway that we're all going to benefit from, which we already are, is we have free intelligence in our pocket.”

When intelligence costs nothing at the baseline, raw API access stops commanding high enterprise multiples. The money pools around whichever company owns the customer's direct attention or solves an end-to-end operational bottleneck.

Why Meta Outlasts Standalone Models

Between the technology giants building frontier models, Shkreli singles out Meta as the company best positioned to maintain high returns. Chatbots are a starting point, not a destination. Current conversational windows will soon look primitive.

"I think it's Meta," Shkreli notes. “They have the distribution with Muse. They showed that they still have product chops. I think we're still in day one of AI. I actually think that the chat bots are this, we're going to look back and laugh and say these are the early days.”

Distribution beats raw model superiority once open-weight software runs locally on consumer devices. Meta does not need to sell access to its weights to win. It monetizes user engagement across billions of existing feeds, turning free intelligence into an engagement wedge while rivals bleed cash running inference farms.

Altman's Hardware Hedge and the In-House S&P 500

If intelligence is destined to become a free utility, pure-play labs face margin compression. OpenAI cannot rely solely on monthly software subscriptions to sustain hundreds of billions in capital expenditure. Shkreli argues that Sam Altman already understands this trap, which is why OpenAI is pursuing physical hardware devices alongside vertical corporate entities.

“Sam's hedge with hardware is really brilliant because he'll always have the best model I think, but being able to deliver that model is probably more valuable,” Shkreli explains. “You're going to have AI starting businesses, and that's I think part of his plan: to have within OpenAI, OpenAI Pharma, OpenAI Motors, OpenAI Aviation, and he'll have his versions of those. You're going to have a whole S&P 500 within OpenAI or Anthropic.”

At the enterprise scale, compute acts as direct capital allocation. “McDonald's can afford to spend $10 billion and say, 'OpenAI, give me 10 million GPUs. I want to know what the best next sandwich we should launch is.' And that's an NPV positive if the intelligence is worth it.” The money goes to domain answers, not raw text generation.

What to Do With This

Audit your product roadmap this week and strip out features that merely wrap raw model prompts. If your business depends on selling an intelligence wrapper that an open-weight model will replicate for free by next year, kill it. Rebuild your distribution channel around proprietary data workflows or vertical operational outcomes that an off-the-shelf chatbot cannot access.