Chamath Palihapitiya went on the Moonshots podcast last week and arrived with a number already in hand: $5 trillion. That is what he says Tesla will be worth once robotaxis own the roads and human drivers become a footnote. The problem is, that number is not his — and it is not even the number he used.
Cathie Wood and ARK Invest called it in April 2023. Wood predicted Tesla would hit $5 trillion by 2027 as its robotaxi fleet rolled out — a forecast that required Tesla to execute flawlessly on autonomy while building an entirely new revenue model dependent on regulatory approval, massive infrastructure investment, and technology that did not yet exist at scale. The $5 trillion figure became shorthand for the thesis, repeated across financial media until it hardened into something resembling consensus among Tesla bulls. Chamath did not originate it. He endorsed it. On the All-In podcast in January, when he floated the reverse merger idea, his own number was $3 trillion — not $5 trillion. The extra $2 trillion is adoption of Wood thesis, not independent calculation.
What Chamath actually brought to the conversation was a different bet. On All-In, he floated an idea that Wood never made: that SpaceX would not pursue a traditional IPO at all, but would reverse-merge into Tesla, consolidating Musk two most valuable assets — SpaceX at roughly $1 trillion and xAI at $250 billion — under a single cap table. Chamath called it the contrarian bet for 2026. The SpaceX-xAI merger closed in February at $1.25 trillion. That part looks prescient.
But the rest of the thesis did not survive contact with reality. Musk confirmed SpaceX is still targeting an IPO at valuations up to $1.5 trillion, despite Chamath prediction that the reverse merger would eliminate that option. Meanwhile, Tesla core auto business declined for the second consecutive year in 2025, with deliveries falling 8.6 percent to 1.64 million vehicles — the steepest annual drop in company history — and Europe described internally as a bloodbath. China posted its first ever annual sales decline. BYD reclaimed the global EV crown.
Chamath is doubling down anyway. On Moonshots, recorded March 23 and published March 26, he laid out a vision of Tesla not merely as a car company or an AI company, but as a vertically integrated compute platform — AI inference in vehicles, humanoid robots, orbital compute, the whole stack. He has said he plans to buy and deploy Cybercab fleets around the country and expects the investment to pay back with positive free cash flow in under two years.
The load-bearing claim is TeraFab.
Musk announced the joint venture on March 21 in Austin, Texas. The plan: a $20-25 billion chip fabrication facility next to Giga Texas, producing 1 terawatt of AI compute annually — roughly 50 times current global AI chip output. Initial targets call for 100,000 wafer starts per month, scaling to 1 million. Small-batch AI5 production is planned for 2026, volume production for 2027. Musk has said 80 percent of TeraFab compute output would be directed toward SpaceX orbital AI satellites, with 20 percent for ground-based applications.
The 50-times figure rests on a baseline Musk himself provided at the announcement: current global AI compute output is approximately 20 gigawatts per year. That number comes from Musk announcement speech — it is a self-cite, not an independent estimate. Independent analysts have not verified the 20 GW figure, and semiconductor industry estimates of current installed capacity vary depending on what counts as AI compute.
The scale question is where the thesis runs into gravity. Bernstein estimates achieving 1 terawatt per year would require $5 trillion to $13 trillion in capital spending and 140 to 360 new factories. For context, TSMC spent $165 billion over years building six Arizona fabs — and those will not reach 2nm production until 2029. Tesla has zero semiconductor fabrication experience. The AI5 chip has already been delayed once, to mid-2027. Samsung 2nm production issues have pushed the AI6 chip back roughly six months.
There is also the xAI financials to consider. The AI company at the center of this vertical integration strategy burned approximately $9.5 billion in the first nine months of 2025 alone, according to The Information. That is the burn rate of a company building hyperscaler infrastructure, not a startup finding product-market fit.
Electrek drew the obvious parallel: this is Battery Day on a larger scale. In 2020, Musk promised Tesla would ramp to 3 terawatt-hours of battery production by 2030. Today Tesla is estimated to be at about 2 percent of that original target. The dry electrode process needed multiple revisions and took years longer than promised.
Musk frames the skepticism as people who did not believe in electric cars or reusable rockets either. That is a reasonable counter to the they have never done it before objection. But chip fabrication at scale is categorically different from anything in Tesla manufacturing history. The capital requirements, technical complexity, and time horizon are in a different category from rockets or batteries.
The $5 trillion number is Cathie Wood. The reverse merger was Chamath, and he got the direction right — but Musk merged xAI into SpaceX, not SpaceX into Tesla, and the IPO is still happening. TeraFab is the bridge between the thesis and reality — and right now, that bridge does not exist.
