E118: AI FOMO frenzy, macro update, Fox vs Dominion, US vs China & more with Brad Gerstner

The prediction bundles two linked claims:

1. Fed funds rate ≈5.5%.
   The Federal Reserve did in fact raise the federal funds target range to 5.25%–5.50% at its July 26, 2023 meeting, the highest level in over 22 years. This was the peak of the 2022–2023 tightening cycle and was maintained into mid‑2024 before cuts began. (cnbc.com)
   On this part, Chamath’s call for a ~5.5% fed funds rate was essentially correct.

2. Credit Suisse 3‑month T‑bill product at ~7.5% soon.
   Three‑month U.S. Treasury bill yields in 2023 peaked around the mid‑5% range, not anywhere near 7.5%. For example, data and commentary show 3‑month T‑bill yields around 5.3%–5.4% in 2023–2024, including specific observations such as 5.37% in May 2023 and 5.34% on October 5, 2023. (nasdaq.com)
   A bank structuring a simple 3‑month T‑bill product cannot sustainably advertise a 7.5% annualized yield when the underlying risk‑free 3‑month bills only yield ~5.3–5.4%; there is no evidence of such an offer from Credit Suisse.

Moreover, Credit Suisse failed and was taken over by UBS in an emergency rescue in March 2023, with the legal takeover completed June 12, 2023. After that point it no longer operated as an independent bank that could roll out new CS‑branded retail products. (cnbc.com) This makes the notion that “soon” in the same cycle Credit Suisse would be offering 7.5% on 3‑month T‑bills even less plausible.

Because the prediction, as normalized, explicitly ties a correct macro call (Fed to ~5.5%) to a much higher, specific yield level (Credit Suisse 3‑month T‑bill product at ~7.5% in the same hiking cycle) that clearly did not occur, the combined prediction must be judged wrong overall, even though one component was accurate.

Evidence from 2H 2023 and 2024 strongly supports Sacks’s forecast of (a) widespread down/structured rounds for 2021‑vintage startups and (b) a bifurcated market where AI attracts most new capital.

1. Down rounds, structure, and recapitalizations for 2021‑era deals

• Carta and PitchBook data show that from early 2023 onward, the share of flat or down rounds in the U.S. stayed near post‑2022 highs, with about a quarter of all Q1 2024 financings globally at flat or reduced valuations and U.S. flat/down rounds above 26%—the highest proportion in more than a decade. (zephyrnet.com)
• Preqin found that down rounds jumped to an unprecedented 27% of all global venture deals in Q4 2023, up from 14.5% the prior quarter, confirming that by late 2023 a large share of financings were already repricings. (linkedin.com)
• Legal/market overviews note that in 2024 nearly 25% of U.S. VC deals and 18% of European deals were down rounds, and investors often imposed “investor‑friendly” terms such as warrants, convertibles, and other structured instruments to take advantage of falling valuations. (practiceguides.chambers.com)
• VC commentary aimed at founders explicitly says that “structured rounds…returned in full force over the past two years as late‑stage companies reel from the heights of 2021,” and that companies mispriced in 2021 face a choice between down rounds or structured rescue financing. (1984.vc)
• A Financial Times analysis of U.S. startups in 2024 describes a 60% jump in failures as many ran out of cash from the 2021‑22 boom, while new funding became scarce except in favored sectors—clear evidence of the “medicine being taken” for peak‑era valuations. (ft.com)

2. AI as the “hot area” vs. everyone else (bifurcated market)

• Global VC investment in 2024 was roughly flat year‑on‑year and still 55% below the 2021 peak, but AI funding grew more than 80% and accounted for close to one‑third of all venture dollars, driven by mega‑rounds for xAI, Databricks, Anthropic, OpenAI and others. (barrons.com)
• Analyses of 2024–25 VC flows show that AI represented only a small fraction of total deal count (e.g., ~3% of U.S. rounds in 2024) but 15% or more of invested capital, with “enormous Series Bs at punchy valuations” for AI while non‑AI startups “fight for any remaining capital.” (thatwastheweek.com)
• Multiple industry pieces explicitly describe the situation as a “tale of two cities”: AI funding surged (e.g., +62% to $110B in 2024) while overall startup funding fell double digits; sectors like fintech and climate tech saw large declines and struggled to raise. (esinli.com)
• Data from Kruze and others indicate that by 2024 a large majority of new capital for many startup portfolios was going into AI, even as failures and shutdowns climbed among non‑AI companies that had raised aggressively in 2021–22. (ft.com)

Taken together, the data from 2H 2023 through 2024 show (1) historically high levels of down and structured rounds affecting many startups that raised at 2021 peak valuations, and (2) sustained, outsized investor enthusiasm for AI deals while most other startups faced a harsh funding environment. That closely matches Sacks’s prediction of “a lot of medicine,” widespread down/structured rounds, and a bifurcated market where AI remained the clear funding magnet.

Evidence strongly supports the first part of Jason’s prediction (a ~10x expansion of generative‑AI startups), but it is too early and too data‑poor to judge the second part (that most of those startups will be rapidly culled by milestone‑based funding).

1. Did the number of generative‑AI startups grow ~10x from ~500 to around 5,000?
Multiple datasets show that, by 2024–2025, the generative‑AI startup population was already well into the thousands:

• StartUs Insights’ 2024 Generative AI report finds 8,600+ generative‑AI companies globally, including 2,037 startups; that’s already far above Jason’s starting premise of ~500. (startus-insights.com)
• Its 2025 report shows continued rapid growth to 16,520 companies and 6,020+ generative‑AI startups worldwide. (startus-insights.com)
• Another analysis of the generative‑AI landscape reports roughly 50,000 generative‑AI firms at end‑2023, rising to about 67,200 by early 2024, indicating explosive proliferation well beyond the 5,000 mark (definitions differ, but all show orders‑of‑magnitude growth vs. 2022–early 2023). (seo.goover.ai)

Taken together, these sources make it quite clear that by 2024–2025 the number of generative‑AI startups globally had very likely exceeded Jason’s 5,000‑startup threshold, so the growth/"10x" portion of the prediction looks directionally right.

2. Has the sector already “quickly whittled down,” with most of those startups culled?
Here the data are much weaker and mostly indirect:

• Funding into generative‑AI startups remains extremely strong: investment jumped from about $25.2B in 2023 to roughly $56B across 885 deals in 2024, a 92% YoY increase—more a capital surge than a post‑bubble culling. (fliphtml5.com)
• However, capital is becoming concentrated: in 1H 2025, AI startups were 53% of global VC funding (64% in the U.S.), yet just five U.S. companies captured over one‑third of Q2 venture dollars, suggesting many smaller startups are starved for follow‑on funding even as headline investment grows. (axios.com)
• S&P Global’s Generative AI Market Monitor shows the top eight vendors now hold ~63% of generative‑AI software revenue, but the number of vendors with >$10M in revenue also increased from 78 to 138 between mid‑2024 and 2025, implying a widening tier of meaningful players rather than an obvious collapse in counts. (spglobal.com)
• An MIT‑linked study finds that 95% of generative‑AI projects in businesses are failing to produce meaningful outcomes, indicating high project‑level failure and likely pressure on many vendors, but it does not give hard data on how many generative‑AI startups have actually shut down. (timesofindia.indiatimes.com)
• Press coverage documents some AI‑startup shutdowns (e.g., several Indian AI startups recently closing), but these are illustrative cases, not evidence that a majority of gen‑AI startups globally have already disappeared. (economictimes.indiatimes.com)

In other words, we do see early signs of the shake‑out Jason anticipated—capital concentrating in a relatively small set of winners, many pilots failing, and some startups shutting down—but the available data:

• show net growth in the number of generative‑AI startups through 2024–2025, not a drop; and
• do not yet let us quantify whether “the majority” of the expanded startup cohort (≈5,000+ companies) has actually been culled.

Given that Jason spoke about growth “over the subsequent few years” (from March 2023) and a quick culling via milestone‑based funding, we are still within that multi‑year window (late 2025). Startup survival/failure statistics at this granularity typically only become clear over a 5‑ to 10‑year horizon. At this point, we can confidently say the 10x‑growth part has happened, but we cannot yet verify or falsify the claim that most of those startups will be rapidly whittled down.

Conclusion: Because the prediction is compound and its second, crucial clause (rapid majority culling) cannot yet be empirically assessed with the data available as of November 2025, the fairest overall verdict is “inconclusive (too early)” rather than fully right or fully wrong.

Salesforce did exactly what Jason described and then went even further.

• Program increased to ~$20B: In February 2023, shortly around the time of the March 2023 earnings discussion, Salesforce’s board authorized an additional $10 billion in share repurchases on top of the existing $10 billion program, for an aggregate total authorization of $20 billion. This is disclosed in Salesforce’s 2023 Form 10‑K and related 10‑Q filings, which state that the August 2022 $10B authorization was expanded by $10B in February 2023 to reach $20B in total. (sec.gov)
• Ongoing, large‑scale buybacks: Salesforce continued to deploy significant capital to repurchases after that point. In fiscal 2023, 2024, and 2025 it repurchased roughly $4.0B, $7.7B, and $7.76B of stock, respectively, as reported in its 10‑K/10‑Q disclosures. (streetinsider.com) The quarterly 10‑Q for the period ended July 31, 2023 also shows multiple billions repurchased and many billions still available under the program, consistent with “seeing more of that for sure.” (annual-statements.com)
• Further expansions beyond $20B (supporting the general thrust): In February 2024, the board authorized another $10B, taking the total authorization from $20B to $30B. (streetinsider.com) Then, in September 2025, Salesforce announced a further $20B increase to its buyback authorization, bringing the total to $50B, as noted in its Form 8‑K summary and contemporary news coverage. (perivis.com)

Given that Salesforce both (a) increased its share repurchase authorization to about $20B as forecast and (b) continued to deploy substantial capital into buybacks over the subsequent years, Jason’s prediction is right.

By late 2025, TikTok’s original, China‑controlled corporate structure has effectively been barred from operating in the U.S., and the app has only survived by undergoing a forced restructuring, which aligns closely with the normalized prediction.

Key facts:

1. Targeted ban on the China‑controlled structure: In April 2024, Congress passed and President Biden signed the Protecting Americans from Foreign Adversary Controlled Applications Act (PAFACA). The law explicitly names ByteDance and TikTok as “foreign adversary controlled” and requires them either to divest from Chinese control or face a nationwide ban on hosting and distributing the app in the U.S. by January 19, 2025. (en.wikipedia.org) TikTok and ByteDance challenged this law in TikTok, Inc. v. Garland, but the Supreme Court ultimately left the law in force, with a federal appeals court and then the Court upholding its constitutionality as a national‑security measure. (techcrunch.com)

2. Legal “ban” on TikTok in its old form, even as enforcement was delayed: Coverage summarizing the situation notes that TikTok has technically been “banned in the U.S. since January 19” under the statute, with app stores supposed to stop distributing it, but successive presidential orders repeatedly delayed enforcement while a divestiture deal was negotiated. (macrumors.com) In other words, the original, ByteDance‑controlled structure is legally prohibited from continuing, even if the government temporarily paused actually pulling the app.

3. Forced restructuring to remove Chinese control: On September 25, 2025, the White House issued an executive order titled “Saving TikTok While Protecting National Security.” It formally determined that a “qualified divestiture” would move TikTok’s U.S. operations into a new U.S.-based joint venture, majority‑owned and controlled by U.S. investors, with ByteDance and other “foreign adversary” entities capped at under 20% ownership. The order emphasizes that the new joint venture, not ByteDance, would control the algorithms, code, content‑moderation decisions, and U.S. user data, which must be stored in an American‑run cloud and subject to intensive U.S. security monitoring. (whitehouse.gov) Reporting around the same time describes ByteDance divesting about 80% of its U.S. TikTok assets to a consortium led by Oracle and Silver Lake, with the algorithm retrained and monitored under U.S. supervision. (reuters.com)

4. Outcome vs. the normalized prediction:

   • The normalized prediction states that within a few years after March 2023, TikTok will be banned from operating in the United States in its then‑current, China‑controlled corporate structure. By January 2025–September 2025, U.S. law and executive action do exactly that: the ByteDance‑controlled configuration is legally prohibited; TikTok can continue only if ownership and operational control are transferred to a U.S.-controlled entity.
   • While the app itself ultimately remains available to users, it does so only after becoming the central test case and main corporate target in U.S.–China tech and data‑security tensions, with a bespoke statute naming it and a forced divestiture regime built around it—precisely the kind of “most obvious victim” Chamath was describing. (en.wikipedia.org)

Because (a) the China‑controlled structure has indeed been effectively banned in law and practice, (b) TikTok was forced into a major, U.S.-controlled restructuring within roughly 2.5 years of the March 2023 prediction, and (c) it became the signature corporate flashpoint in U.S.–China tech tensions, the prediction—under the user’s normalized wording—is best classified as right, even though the government ultimately chose forced divestiture rather than a permanent shutdown of the app for users.

Sacks’ normalized prediction was that the Dominion Voting Systems defamation lawsuit against Fox News would not ultimately result in Fox being found legally liable under the New York Times v. Sullivan standard—i.e., Fox would avoid an adverse final judgment on liability.

What actually happened:

• On March 31, 2023, the Delaware trial judge granted partial summary judgment for Dominion, ruling that Fox’s statements about Dominion were false and defamatory as a matter of law, but he left the key actual malice and damages questions for a jury at trial. This meant there was not yet a final judgment of liability under the full defamation standard. (en.wikipedia.org)
• Jury selection began in April 2023, and trial was set to start on April 18, 2023. Just as opening statements were about to begin, the parties announced a settlement: Fox agreed to pay Dominion $787.5 million, and the trial was canceled. (washingtonpost.com)
• Because the case settled, there was no jury verdict and no final court judgment finding Fox liable under the New York Times v. Sullivan “actual malice” standard; the case was resolved by settlement and then dismissed. (en.wikipedia.org)

Although Fox paid a massive sum and the court formally found that many of its statements about Dominion were false, the specific prediction was about avoiding an adverse final liability judgment under the governing defamation standard. That is exactly what occurred. Therefore, judged against the normalized prediction, the prediction is right.

Evidence since 2023 shows that TikTok has, in practice, become a major casualty of escalating U.S.–China great power competition, primarily via a forced divestiture rather than an outright ban.

Key developments:

1. Congress targeted TikTok directly with a divestiture-or-ban law. In April 2024, the Protecting Americans from Foreign Adversary Controlled Applications Act was signed into law. It explicitly applies to ByteDance/TikTok and requires apps deemed “foreign adversary controlled” (i.e., TikTok) to either be divested from that control or be banned in the U.S. by January 19, 2025, with only a short possible extension. (en.wikipedia.org) This law is a direct outgrowth of U.S. national security concerns about China.

2. Courts upheld the law, making clear TikTok faced an existential choice. A federal appeals court in late 2024 upheld the constitutionality of the law, confirming that TikTok must complete a qualified divestiture from ByteDance by January 19, 2025, or effectively become unavailable in the U.S. (jurist.org) In early 2025, the U.S. Supreme Court likewise upheld the statute in TikTok, Inc. v. Garland, affirming that Congress could mandate divestiture because of national security risks tied to Chinese ownership. (americanbar.org)

3. The U.S. government repeatedly delayed but did not back off the mandate. After taking office in January 2025, President Trump issued a series of executive orders delaying enforcement of the TikTok ban provisions, not to spare TikTok indefinitely, but explicitly to allow time to negotiate and complete a sale of TikTok’s U.S. operations that would satisfy the statute’s requirements. (whitehouse.gov) The White House framed this as balancing national security with the desire to keep the app online.

4. By late 2025, ByteDance is being forced to relinquish control of TikTok U.S. A September 25, 2025 executive order from President Trump found that a proposed transaction to move TikTok’s U.S. app into a new U.S.-based joint venture constitutes a “qualified divestiture” under the Act, with ByteDance’s stake to be kept under 20% and the algorithm and content moderation controlled by the new U.S. entity under heavy security monitoring. (whitehouse.gov) Reporting in November 2025 describes ByteDance as being in the process of divesting about 80% of its U.S. assets in TikTok to a consortium led by Oracle and Silver Lake, with TikTok’s algorithm to be retrained and supervised under U.S. oversight. (reuters.com)

Interpretation against the prediction:

• The user’s normalized version of the prediction allows "ban, forced divestiture, or equivalent measures" as the mechanism by which TikTok is “severely harmed” and becomes “GPC roadkill.”
• While TikTok as a product remains available in the U.S. as of late 2025, its China-based owner is being compelled by U.S. law and policy—explicitly justified by national security concerns about China—to surrender majority ownership and operational control, including over its core algorithm and data handling. That is a severe strategic and economic blow to ByteDance and to TikTok as a China-controlled asset, and it is precisely the type of outcome one would describe as a major casualty of great power competition.
• The timeline (“over the few years following March 2023”) is also satisfied: the law was passed in April 2024, upheld through 2024–2025, and by late 2025 the forced divestiture framework is in place and being executed.

Because U.S.–China competition has in fact led to a binding legal framework that forces TikTok’s U.S. operations out of effective Chinese control under threat of a nationwide ban, within roughly 2.5 years of the March 2023 podcast, the core substance of Sacks’s prediction—that TikTok would end up as GPC “roadkill” via ban/forced divestiture or equivalent severe measures—is best judged as right rather than wrong or ambiguous.

Evidence since 2023 shows that competition with China framed as great/strategic power competition has in fact functioned as a primary organizing principle of U.S. foreign policy and strategy.

1. China-centered competition as core of official strategy. The 2022 National Security Strategy (still operative through the following years) explicitly states that a competition among major powers is underway and identifies China as “America’s most consequential geopolitical challenge,” the only state with both the intent and capability to reshape the international order, and says the U.S. must “out-compete” China. It describes the era as one of “strategic competition with major powers,” clearly elevating this competition—especially with China—above other threats.【2†turn2search1】【2†turn2search2】 Independent summaries stress that this competition with China suffuses the document.【2†turn2search0】【2†turn2search4】

2. Defense planning built around China as the pacing challenge. The unclassified 2022 National Defense Strategy and related Pentagon statements repeatedly call China the “overall pacing challenge” and “most consequential strategic competitor for the coming decades,” saying U.S. force planning, modernization, and posture are set with that yardstick in mind.【0†turn0search2】【0†turn0search5】 Analyses of this framing note that U.S. defense planning, industrial policy, and global posture now revolve around how to keep up with and outpace Beijing, explicitly treating China-centric competition as the benchmark for strategy.【0†turn0search9】

3. Congressional and bureaucratic structures explicitly organized around competition with China. In January 2023, the House created the Select Committee on the Strategic Competition between the United States and the Chinese Communist Party, mandated to study and make recommendations across economic, technological, and security domains of U.S.–China rivalry.【2†turn2search14】 That a standing select committee is framed entirely around "strategic competition" with China is strong evidence that this concept has become a central organizing lens in official policy deliberations.

4. Foreign-policy professionals explicitly describe GPC as the central organizing principle. A U.S. Foreign Service Journal article (June 2021, still widely cited in subsequent debates) flatly describes the re-emergence of “great power competition” as “the central organizing principle for U.S. engagement with the world,” specifically highlighting the rise of Xi’s China as the main peer competitor.【4†turn4view0】 Later commentary and book reviews on U.S. strategy in the era of “great power competition” and “strategic competition” treat this frame as the widely accepted deep structure of U.S. grand strategy, even when arguing it is a flawed or dangerous organizing concept.【3†turn3search0】【3†turn3search4】

5. Ongoing China-focused competitive framing across regions and issues. Analyses of U.S. policy in regions like the Middle East complain that great power competition with China is increasingly the lens through which Washington views local issues—implicitly acknowledging that this frame is driving policy.【1†turn1search10】 Other scholarship and commentary in 2023–2025 describe the United States as entering a full-fledged "great-power contest" or “great-power rivalry” with China that dominates policymakers’ thinking about alliances, technology controls (e.g., chips), and economic statecraft.【1†turn1search1】【2†turn2news12】【2†turn2search6】

6. Caveats that do not overturn the prediction.

   • The Biden team nominally shifted terminology from “great power competition” to “strategic competition” in some documents,【3†turn3search3】【3†turn3search13】 and the NSS also emphasizes transnational threats (climate, pandemics) alongside competition.【2†turn2search3】
   • Some 2024–2025 analyses argue that the great-power-competition frame is misguided or even in decline under newer approaches (e.g., more homeland-focused defense, spheres-of-influence bargaining).【0†turn0news16】【1†turn1search11】【1†turn1search9】

   But these critiques presuppose that great/strategic power competition with China is the prevailing paradigm they are arguing against, which reinforces—rather than refutes—Sacks’s claim that it has become the organizing principle.

Taking all of this together: by the mid‑2020s, China-centered great/strategic power competition clearly shapes U.S. national security strategy, defense planning, congressional structures, and much of the foreign‑policy debate. That matches Sacks’s prediction that this concept would become the primary organizing framework of U.S. foreign policy over the years after March 2023, even if the preferred official label is often “strategic competition” rather than the exact phrase “great power competition.”