DeepSeek Panic, US vs China, OpenAI $40B?, and Doge Delivers with Travis Kalanick and David Sacks

As of November 30, 2025, only about ten months have passed since the prediction, which referred to changes happening “over the next several years,” so the full time horizon has not elapsed. Current evidence is mixed: on one hand, proprietary frontier‑model companies remain extremely valuable and clearly not “worth nothing.” OpenAI completed a secondary share sale in 2025 valuing it at around $500B, and reporting indicates that sale closed in October 2025 at that valuation. (cnbc.com) Anthropic has raised successive multi‑billion‑dollar rounds, with valuations climbing from about $61.5B in March 2025 to roughly $183B by September 2025, alongside rapidly growing run‑rate revenue and hundreds of thousands of business customers, which strongly suggests investors still view its proprietary models as highly differentiated assets. (anthropic.com) On the other hand, there is very rapid progress and adoption of powerful open or open‑weight models: Meta’s Llama 3 and later the 405B‑parameter Llama 3.1 are released under relatively permissive licenses, positioned as “open” models and advertised as competitive with leading proprietary systems on many benchmarks. (techcrunch.com) DeepSeek‑R1 and related Chinese open models are fully open‑source (e.g., MIT‑licensed), show reasoning performance competitive with GPT‑4‑class models on math, medical, and other benchmarks, and are distributed via GitHub and Hugging Face at low or zero license cost, contributing to a surge in downloads of “open” models where China now slightly leads the U.S. (arxiv.org) Meanwhile, commoditized access to models via cloud APIs is growing, but the market behavior doesn’t yet look like pure undifferentiated commodity infrastructure: AWS Bedrock, for example, has faced AI capacity shortages that pushed customers to rivals, indicating that access to top‑tier model capacity remains scarce, and leading vendors like OpenAI and Anthropic still command premium per‑token prices for their highest‑end models compared with cheaper, lower‑tier offerings. (businessinsider.com) Taken together, this suggests that Jason’s strong present‑tense claim in early 2025 (that frontier LLMs have already been “proven” to be worthless as proprietary assets and are all headed to open‑source/commodity status) is not supported by current facts; however, because his normalized prediction is explicitly about what will happen over several years, it is too early to determine definitively whether the longer‑run commoditization he describes will ultimately occur, so the prediction must be rated as inconclusive at this time.

Available evidence shows Jason’s quantitative scenario has already failed and cannot be salvaged, even though Trump’s second term is not over.

1. Baseline of ~$1B/day and goal of >$3B/day
   DOGE’s official X account and sympathetic coverage in January 2025 said it was already “saving the Federal Government approx. $1 billion/day” and explicitly framed > $3 billion/day as a future target, not an achieved rate.(foxbusiness.com) There is no later official DOGE communication or mainstream report indicating that DOGE ever claimed to have reached a sustained $3B/day savings rate.

2. Actual and claimed totals are far below what 3B/day for years would imply
   To get roughly $60,000 in implied benefit for a family of five, Jason was implicitly extrapolating something like $3B/day over most of a 4‑year term, i.e., on the order of $4+ trillion in total savings/claims (about $13k+ per person, or ~$65k for five people).
   By contrast, DOGE’s own official Savings page listed $214B in estimated savings as of October 4, 2025, with an “amount saved per taxpayer” of $1,329.19.(doge.gov) Even if you simply scale that per‑taxpayer figure by five, you get under $7,000 for a family of five—roughly an order of magnitude below the ~$60,000 Jason projected, and that’s after most of 2025 when DOGE was active.

Independent and semi‑official analyses are even less generous. A Reuters deep dive in May 2025 found that while DOGE claimed roughly $175B in savings, the actually observable reduction in agency spending over comparable periods was only about $19B, i.e., roughly half of one percent of federal outlays—nowhere near multi‑trillion‑dollar territory.(reuters.com) The Wikipedia summary of Trump’s second‑term economic policy similarly notes that Elon Musk revised DOGE’s goal down from $2T to about $150B in savings and that, by April 2025, DOGE was claiming a bit over $160B, with much of that offset by the costs of mass layoffs and disruptions.(en.wikipedia.org) None of these trajectories point toward trillions in claimed savings.

3. DOGE has already been dissolved, foreclosing a later catch‑up to 3B/day
   DOGE was created by Trump’s executive order as a temporary organization within the Executive Office of the President, with a mandate initially running into mid‑2026.(en.wikipedia.org) But in late November 2025, multiple outlets reported that DOGE had been quietly disbanded about eight months ahead of schedule; its functions were partly absorbed by the Office of Personnel Management.(theguardian.com) Reuters and others describe it as no longer functioning as a central entity.(reuters.com) Once the department itself has been shuttered, it can no longer ramp up to or sustain a new $3B/day claimed savings rate over the remainder of Trump’s term.

4. Macro indicators contradict the idea of large net savings
   Separate analyses of federal fiscal data show that, despite DOGE’s rhetoric, overall federal spending and debt continued to rise, with spending in early 2025 running notably higher than under the prior administration.(marketwatch.com) While Jason only talked about claimed savings, the fact that neither DOGE’s official tallies nor macro‑level fiscal aggregates ever approach the multi‑trillion‑dollar scale implied by his back‑of‑the‑envelope math reinforces that his scenario did not materialize.

Putting it together: DOGE peaked at a (contested) claim of about $1B/day and a few hundred billion in total claimed savings before being shut down; it never reached or sustained $3B/day, and the cumulative implied benefit per family of five is an order of magnitude below the ~$60k Jason projected. With DOGE now dissolved, there is no realistic path for the original prediction to become true. Therefore this prediction is wrong.

Trump's Department of Government Efficiency (DOGE) was created on Jan 20, 2025 as a temporary initiative in his second administration to cut federal waste and spending. (en.wikipedia.org) Trump's second term is scheduled to run until Jan 20, 2029, so the end of Trump's second term has not yet occurred as of Nov 30, 2025. (en.wikipedia.org) DOGE itself was intended to last until July 4, 2026 but was quietly dissolved in November 2025, with remaining functions absorbed into the Office of Personnel Management. (reuters.com) Publicly reported DOGE savings to date are at most in the hundreds of billions of dollars claimed by its proponents, and independent audits suggest only a few billion in verifiable cuts while overall federal spending and deficits have continued to rise; there is no credible evidence of more than $2 trillion per year in annualized reductions identified by DOGE so far. (reuters.com) However, because Trump’s second term does not end until 2029, the time horizon specified in Chamath’s prediction has not yet elapsed, so the prediction cannot be definitively judged right or wrong at this point.

Friedberg forecast three linked steps: (1) an aggressive wave of DOGE/Trump spending cuts within the next few months, (2) court fights that clarify what the executive can and cannot cut unilaterally, and (3) a shift toward pushing cuts through Congress.

1. Whirlwind of cutting: After Trump’s January 20, 2025 order creating DOGE, the administration rapidly rolled out mass layoffs, workforce‑optimization orders, contract cancellations and grant terminations across many agencies in February–April 2025, including large reductions at HHS, Education, HUD, VA, IRS and others. (en.wikipedia.org) That fits a fast, aggressive attempt to cut spending of exactly the kind he described.

2. Courts adjudicating limits: Starting on January 20, 2025, dozens of lawsuits challenged DOGE and related executive orders for impoundment of appropriated funds, dismantling USAID, and large‑scale reductions in force. (en.wikipedia.org) On May 9 and May 30, 2025, the AFGE v. Trump case produced a TRO and then a Ninth Circuit ruling that the president lacks authority to direct broad government‑wide RIFs and reorganization without congressional authorization, explicitly drawing a line between what can be done by executive order and what requires Congress. (clearinghouse.net)

3. Turn to Congress: As these legal concerns mounted, reporting describes the White House preparing formal rescission requests under the Impoundment Control Act and sending Congress a package to cancel roughly $9.3 billion in funding for PBS, NPR and State Department accounts, and then advancing the One Big Beautiful Bill Act with over a trillion dollars in targeted spending cuts through the House beginning in late April and May 2025. (cbpp.org)

Taken together, within a few months of the January 31 episode there was an aggressive cutting drive, major court rulings on its legality, and an ensuing push for legislative cuts, matching his prediction in substance and timing.

Using the Federal Reserve’s official DGS30 series (30‑year constant‑maturity Treasury, daily), the yield on January 31, 2025 was 4.76%. Over the subsequent months it rose but peaked at 5.08% on May 22, 2025, never reaching the ~5.5% level Friedberg relayed. After that peak, the yield trended lower, standing around 4.64% by November 26, 2025, well below both the 5%+ highs and the 5.5% target. (fred.stlouisfed.org)

Contemporaneous market coverage describes this same move as the 30‑year yield surging a bit above 5%—around 5.08–5.09%—its highest close since 2023, again with no mention of it approaching 5.5%. (barrons.com) Because the cycle has clearly seen a rise and then a material decline without ever touching ~5.5%, Friedberg’s quoted prediction that the 30‑year yield would go to about 5.5% “before it goes down” did not come true.