Meta's scorched-earth open-source AI strategy and the future of Tesla's businesses

Open-sourcing best-in-class models commoditizes foundational AI and preserves the ad-based moat of incumbents.

Meta released Llama 3 as fully open-source, instantly matching GPT-4 quality and forcing VCs to stop funding closed competitors.

Why: By making the base layer free, Meta destroys the economic value of proprietary models and channels innovation into its own ecosystem.

"They're scorching the earth for these new markets so that the economics are not viable"

Founder-led companies with voting control can make long-term, counter-consensus capital allocations.

Zuckerberg and Musk both announced multi-billion-dollar infrastructure spend that public-market investors initially punished.

Why: Concentrated founder authority insulates strategic decisions from short-term shareholder pressure.

"It's a really important sign that founder led companies with these voting rights... have the ability to make these sorts of hard decisions"

Inference compute will be 100× larger than training compute, so overspending on NVIDIA GPUs for inference is misallocated capital.

Market ‘sharps’ sold Meta on the thesis that NVIDIA hardware is poorly suited for inference workloads.

Why: Training demands massive parallel matrix math; inference can be served by cheaper, specialized silicon.

"AI is really two markets, training and inference. And inference is going to be 100 times bigger than training"

Integrating an AI search box inside Instagram, WhatsApp and Facebook gives Meta a path to steal 10 points of global search share.

Meta.ai now sits at the top of three billion-user apps, combining psychographic and intent data no rival possesses.

Why: Each search point is worth ~$150 B in market cap; Meta can fuse social graph data with query intent for unmatched ad targeting.

"Meta is going to get ten points of the search market... ten points is worth 150 billion in market cap"

Tesla’s stock reaction is best understood by tracking ‘sharps vs squares’—smart money rewards adherence to the 2016 Master Plan.

Despite missing earnings, Tesla rallied because every strategic milestone (Model 3, Y, Cybertruck, FSD, robotaxi) is on the public roadmap.

Why: Markets price predictable execution; deviations from stated plans create dislocations that informed capital exploits.

"Sharps love it... he's actually executing exactly the plan... we don't have to actually guess what the plan is because he puts it out on the website"

Robotaxi ride-hailing is Tesla’s largest TAM because it removes the human labor cost that keeps Uber/Lyft prices high.

With FSD 12 already capable of constrained-area autonomy, Tesla can undercut incumbents and capture massive utilization of its fleet.

Why: Eliminating the driver collapses marginal cost per ride and unlocks the original ‘full Uber’ vision of universal on-demand transport.

"Ride-Sharing is number one... by an order of magnitude... if Tesla enters... and just guts pricing... this thing is just gang busters nuclear"

Residential energy (solar + battery) can turn 100 million US homes into mini-utilities, undermining 1,700 legacy utilities.

Regulated utilities must spend $2 T in CapEx and pass costs to ratepayers, while rooftop solar approaches zero marginal generation cost.

Why: Net-metering laws plus cheap panels create a decentralized grid that obsoletes centralized utility debt.

"Make them mini utilities... you're totally resilient and independent of the utilities... short the debt of these utilities"

Banning non-competes accelerates innovation by letting talent, tacit knowledge and best practices flow freely across companies.

California’s prohibition on non-competes is credited for Silicon Valley’s rapid iteration; the FTC now extends that nationwide.

Why: Ideas in employees’ heads become public goods; frictionless mobility raises wages and startup formation.

"One of the reasons why the tech industry moves so fast... is because of the happy coincidence that California doesn't allow non-competes"