E135: Wagner rebels, SCOTUS ends AA, AI M&A, startups gone bad, spacetime warps & more

As of November 30, 2025, Vladimir Putin remains in power as President of Russia. He was inaugurated for his fifth presidential term on May 7, 2024, following the 2024 election; this term is scheduled to run until 2030, and constitutional changes allow him to run again thereafter. (trtafrika.com) He continues to act as head of state in 2025, including signing international treaties such as the 2025 Iran–Russia comprehensive strategic partnership. (en.wikipedia.org)

The prediction’s time horizon is roughly 10 years from July 1, 2023, i.e., until about July 1, 2033. Since that date has not yet been reached—and Putin has neither lost power nor undergone a clearly documented removal driven by illness or the consequences of the Ukraine invasion—the forecast cannot yet be judged as either correct or wrong. It simply remains unresolved at this time.

The prediction was explicitly conditional: if the 2023 Ukrainian counteroffensive were successful "on some level" (i.e., with meaningful territorial gains), then within a few months Russia would move toward or declare full mobilization and escalate to a higher level of violence.

1. Condition not met (counteroffensive outcome):

   • Ukraine’s 2023 counteroffensive, launched in June 2023, recaptured roughly 370 km² and a small number of villages, never reached key operational objectives like Tokmak or the Sea of Azov, and was widely described by late 2023 as having failed to regain significant territory or achieve major strategic goals.(en.wikipedia.org)
   • In other words, the normalized trigger — “a meaningful level of success (e.g., significant territorial gains)” — did not occur.

2. Consequence also did not occur, but without the trigger:

   • Russia had already declared a partial mobilization in September 2022. That decree legally remains in force, but it was not expanded into a declared general/full mobilization after the 2023 offensive.(en.wikipedia.org)
   • Through late 2023 and 2024, Russian officials repeatedly said there were no plans for additional mobilization, instead relying on volunteer contract soldiers, financial incentives, and regular conscription cycles.(english.alarabiya.net)

Because the antecedent of the conditional (“successful” counteroffensive with substantial gains) did not materialize, we have no real-world test of the claim “if that happens, then full mobilization and a higher level of escalation will follow.” The fact that Russia did not fully mobilize after an unsuccessful offensive does not logically falsify that conditional.

So, even though enough time has passed to observe events, the specific if-then forecast cannot be definitively graded as right or wrong. It falls into the category of “ambiguous” rather than clearly correct or incorrect.

Chamath’s prediction was that, after the June 2023 Supreme Court decision ending race-based affirmative action in college admissions, U.S. colleges and universities receiving federal funds would be forced, via lawsuits and legal pressure, to significantly curtail or eliminate both legacy- and athletic-based admissions preferences.

What actually happened by late 2025:

1. Legacy admissions have been reduced, but not eliminated, and mostly not by court rulings.

   • An Education Reform Now (ERN) analysis summarized in Forbes reports that as of 2025 only 24% of four‑year colleges still consider legacy status, down from 49% in 2015, and that 92 colleges dropped legacy preferences after the 2023 affirmative‑action ruling. This shows meaningful decline, but not disappearance. (forbes.com)
   • State legislatures, not federal lawsuits, have been the main source of legal change: by 2024, Colorado, Maryland, Virginia, Illinois, and California had enacted bans on legacy admissions in some or all public institutions, with California’s ban extending to private colleges that take certain state aid. (en.wikipedia.org)
   • These laws are geographically limited and do not cover all federally funded institutions. Many private and out‑of‑state universities that receive substantial federal funding still retain legacy preferences.

2. Major elite universities still use legacy preferences and have not been forced by courts to stop.

   • Harvard remains under a pending civil‑rights complaint and U.S. Department of Education investigation alleging that its legacy and donor preferences violate Title VI, but there has been no final ruling or order requiring Harvard to end legacy admissions as of November 2025. (forbes.com)
   • A 2025 Harvard Crimson piece on state efforts to ban legacy admissions notes that Massachusetts has not passed such a ban and Harvard continues to use legacy preferences, underscoring that even the most scrutinized schools have not yet been compelled to end them. (thecrimson.com)
   • In California, rather than abandon legacy preferences, Stanford chose to withdraw from the state’s Cal Grant financial‑aid program so it could keep giving legacy and donor preference while sidestepping the new state ban—the opposite of being forced to drop legacy by law. (sfchronicle.com)
   • An AP report this year notes that roughly 500 U.S. universities, including all Ivy League schools and Stanford, still consider legacy status, and criticizes the Trump administration for focusing on race-based policies while taking no action on legacy admissions. (apnews.com)
   • Together, these facts show: substantial federal funding continues to flow to many institutions that maintain legacy preferences, with no nationwide legal requirement to end them.

3. There has been legal pressure on legacy admissions, but not decisive wins that make them illegal.

   • Civil‑rights groups filed a Title VI complaint against Harvard’s legacy and donor preferences immediately after the Supreme Court ruling, and the Department of Education opened a formal investigation—clear pressure, but not yet a successful lawsuit that forces policy change. (insightintoacademia.com)
   • State bans in a handful of states are legislative actions, not the kind of federal civil‑rights lawsuit victories Chamath described as the mechanism that would make legacy admissions “go away” at federally funded schools. There is still no Supreme Court or federal appellate decision declaring legacy preferences unlawful per se.

4. Athletic‑based admissions preferences have not been significantly curtailed by law.

   • High‑profile college‑sports litigation since 2023 has focused on NIL compensation, revenue sharing, and transfer rules (e.g., Tennessee v. NCAA and House v. NCAA), leading to major changes in how athletes are paid and how they can move between schools. (en.wikipedia.org)
   • These cases do not challenge or eliminate admissions boosts for recruited athletes; they target economic/antitrust issues, not the practice of admitting athletes with lower academic metrics.
   • There is no evidence of a successful post‑2023 lawsuit that directly attacks athletic‑recruit admissions preferences themselves under Title VI or the Equal Protection Clause, nor of universities being ordered to dismantle athletic preferences as a condition of receiving federal funds.

5. Net assessment versus the prediction.

   • Chamath anticipated that, once race‑based affirmative action was struck down, it would be a “pretty quick” next step for lawsuits to succeed against legacy and athletic preferences, effectively forcing their end at federally funded institutions.
   • By late 2025:
     • Legacy admissions are under intense scrutiny and have shrunk meaningfully, but hundreds of federally funded colleges—especially the most elite—still use them, and there is no controlling court decision or federal rule that bans the practice nationwide. (forbes.com)
     • Athletic‑based admissions preferences remain intact and legally unchallenged in any comparable way; the legal action around college sports has been about money and mobility, not admissions hooks. (en.wikipedia.org)

Because the core of the prediction was that legal victories and resulting mandates would make legacy and athletic preferences “go away” at federally funded schools, and that has clearly not occurred by November 2025, the prediction is best classified as wrong, even though parts of the landscape (especially political and reputational pressure on legacy admissions) moved in a direction broadly consistent with his concerns.

Evidence since mid‑2023 shows that the dynamic Chamath described has in fact played out:

1. Race‑exclusive corporate programs have been sued and forced to change or end.

   • Pfizer’s high‑profile Breakthrough Fellowship Program, originally limited to Black, Latino and Native American applicants, was challenged by the group Do No Harm as unlawful race discrimination. After the Supreme Court’s 2023 affirmative‑action ruling, a U.S. appeals court revived the case; Pfizer then revised the criteria in 2023 to be race‑neutral and, in a 2025 settlement, agreed to stop accepting new fellows altogether. (reuters.com)
   • Edward Blum’s American Alliance for Equal Rights has brought a series of suits against race‑exclusive private‑sector programs (e.g., law‑firm diversity fellowships at Winston & Strawn, Perkins Coie, Morrison & Foerster). At least one, Winston & Strawn, changed its fellowship eligibility to race‑neutral and the case was then dismissed. (washingtonpost.com)
   • The Fearless Fund venture‑capital grant program for Black women business owners, backed by major corporate partners, was held likely unlawful under 42 U.S.C. § 1981 by the Eleventh Circuit, leading to an injunction and later settlement; the race‑exclusive grant program was shut down. (mondaq.com)

2. Broader corporate DEI programs are being rolled back under legal and political pressure.

   • Meta disbanded its DEI programs in early 2025, explicitly citing a changed U.S. legal/policy environment and growing scrutiny of DEI and perceived preferential treatment, and dropped tools like its Diverse Slate hiring approach and representation goals. (theverge.com)
   • Many large companies (e.g., Salesforce, Amazon, Google, Meta, Walmart, McDonald’s, Citi) have scaled back or rebranded DEI initiatives, often softening or removing explicit race‑based targets in response to post‑SFFA legal and political risk. (businessinsider.com)

3. Courts and regulators have made it riskier to rely on explicit racial preferences in employment‑adjacent metrics and programs.

   • In Ames v. Ohio Department of Youth Services (2025), the Supreme Court unanimously removed heightened pleading standards for so‑called “reverse discrimination” Title VII claims, explicitly recognizing that discrimination suits against DEI‑related practices must be evaluated under the same standard as other discrimination claims. Commentators note this will fuel challenges to employer DEI programs post‑SFFA. (en.wikipedia.org)
   • Conservative groups and some state officials now explicitly warn corporations that race‑based hiring, promotion goals, or contracting preferences pose litigation and enforcement risk; shareholder campaigns at Apple and others argue DEI creates significant legal exposure in light of recent Supreme Court decisions, even as Apple has so far chosen not to dismantle its DEI structure. (cnbc.com)

Taken together, these developments match Chamath’s forecast in substance: major private employers and allied institutions running explicitly race‑based hiring, recruiting, fellowship, or grant programs have faced legal challenges that forced them to materially revise or end those programs, and at least some DEI/ESG practices built on explicit racial preferences have been deemed unlawful or untenable in the U.S. Not every big name he mentioned (e.g., Apple, Exxon) has yet lost a case or completely abandoned DEI, but the general prediction about post‑SFFA legal attacks on corporate race‑based initiatives and the resulting pullback has proven accurate.

Chamath’s prediction has broadly played out along both dimensions he named: sequence (higher ed first, then spillover) and scope (from campuses into private enterprise).

1. Higher ed effects emerged first, and grew over time

   • The June 2023 Students for Fair Admissions v. Harvard/UNC decision directly bound colleges and universities, forcing selective schools that had considered race in admissions to redesign processes while most other institutions were less affected.(brookings.edu)
   • In the first admissions cycles after the ruling, analysts noted uncertainty and predicted that the full impact would take years to see, with early reports showing some but not uniform changes in the racial makeup of incoming classes.(csmonitor.com)
   • By 2025, data show clearer and larger shifts: an AP analysis of 20 selective schools found Black freshman enrollment had dropped at nearly all of them, with steep declines at places such as Harvard (from about 18% of the 2023 class to 11.5% of the 2025 class).(apnews.com) This pattern—initial legal/policy adjustments followed by more visible demographic changes over subsequent classes—matches the “slow and then fast” dynamic he described.

2. The first domain was higher education, including later-explicit extensions

   • Immediately and necessarily, the ruling applied to civilian higher-ed admissions. Subsequent litigation sought to extend its logic even to institutions the Court had initially carved out: Students for Fair Admissions sued West Point and the U.S. Air Force Academy over race-conscious admissions, and the Naval Academy ultimately stopped considering race in admissions, explicitly aligning with the Court’s reasoning.(cbsnews.com)
   • This shows the first wave was firmly centered on higher education before other sectors became primary targets.

3. Then similar legal and political pressure spread into private enterprise

   • Conservative legal groups explicitly used the SFFA precedent and related civil-rights statutes to attack private-sector diversity programs. The American Alliance for Equal Rights sued major law firms Morrison Foerster and Perkins Coie over diversity fellowships; both firms revised their programs to remove race-based eligibility, and the suits were then dropped.(washingtonpost.com)
   • The same network of activists brought a high‑profile case against the Fearless Fund’s grants for Black women founders under Section 1981 of the Civil Rights Act of 1866; a federal appeals court blocked the program, and the fund later permanently shut it down as part of a settlement.(wsj.com)
   • Professional and quasi‑corporate programs have also been forced to change: the State Bar of Wisconsin narrowed its diversity clerkship criteria away from race after a discrimination lawsuit, which the bar and commentators linked to the post‑SFFA climate.(reuters.com)
   • Large corporations have rolled back or restructured DEI efforts “amid conservative criticism and legal pressure inspired by the 2023 Supreme Court ruling,” including firms like Walmart and Lowe’s, which reduced race‑focused initiatives and DEI commitments.(apnews.com) Legal and political campaigns now regularly cite SFFA when challenging hiring, fellowships, supplier-diversity and grant programs in the private sector, demonstrating the predicted spillover.

Given the timeline (2023–2025) and the evidence that: (a) higher ed saw the earliest and most direct effects, which became more visible over successive application cycles, and (b) similar legal theories and political pressure are now reshaping DEI and race-conscious programs across private enterprises, Chamath’s forecast that the ruling’s impact would be gradual then rapid in higher education, and then extend into private enterprise policies and programs is substantially borne out by events to date.

Evidence since mid‑2023 shows that the severe shortage of high‑end AI GPUs (Nvidia A100/H100‑class and equivalents) persisted well beyond a year and was still meaningfully constraining the industry into 2025, which matches Sacks’s “year or two” timeline.

Key points:

• In 2023, Nvidia’s AI accelerators were capped by TSMC’s advanced CoWoS packaging capacity; demand vastly exceeded supply. Analyses note that even by early 2024, this packaging bottleneck was only starting to ease, with TSMC planning to more than double CoWoS output by 2025 precisely because supply remained tight relative to AI demand.【(siliconanalysts.com)】
• High‑bandwidth memory (HBM), which is required for modern AI GPUs, has been a separate choke point. Industry analysts and Micron’s CEO have said HBM shortages are expected to weigh on GPU production into 2026, indicating that the underlying constraints did not disappear quickly in 2023–2024 but instead persisted for multiple years.【(luckboxmagazine.com)】
• A concrete example from a hyperscaler: Amazon’s internal documents show its retail division struggled with a lack of GPU capacity throughout 2024, to the point it launched “Project Greenland” in July 2024 to ration and centrally manage scarce GPUs. Reporting in April 2025 says Amazon’s retail business had only then resolved its internal GPU shortage and expected surplus capacity later in 2025—i.e., constraints lasted well past mid‑2024 and into early 2025.【(businessinsider.com)】
• Separate coverage of Nvidia H100 supply notes that through 2024 customers still faced multi‑month lead times; only by later in the year was it plausible that wait times might disappear, and even then large LLM developers like OpenAI were singled out as still hitting supply bottlenecks due to their scale.【(1ai.net)】
• Broad 2023–2025 retrospectives on the “GPU crisis” and AI boom describe a persistent global imbalance where big tech, specialized GPU clouds, and a handful of well‑funded players soaked up most of the Nvidia GPU supply, leaving many startups and smaller firms facing long waits and capacity constraints through 2024 and into 2025.(buzzbyte.org)}

Putting this together: the AI‑GPU shortage clearly did not materially ease within a few months of July 2023; instead, serious constraints persisted for roughly two years, with major players still short into early 2025 and structural bottlenecks (CoWoS, HBM) projected to remain tight even beyond that. That aligns well with Sacks’s claim that the shortage was “probably not going to get better for a year or two,” so the prediction is best judged as right.

Since July 2023, there have been many acquisitions where data infrastructure or enterprise software companies bought AI/tooling firms to stay competitive in the AI market, matching Friedberg’s prediction that “there are more acquisitions still to come.”

Examples include:

• Databricks (a leading data/AI platform) continued buying AI-related tooling companies after MosaicML, including Tabular (a data-management system used by open‑source AI) in 2024 and later Tecton, a machine‑learning/feature‑store startup, in 2025 to strengthen its AI‑agent capabilities. (en.wikipedia.org)
• Snowflake followed its Neeva deal with the acquisition of Crunchy Data, a PostgreSQL startup, explicitly to enhance its ability to help customers build AI agents on their own data. (wsj.com)
• Nvidia, an AI infrastructure leader, acquired OctoAI (formerly OctoML), a generative‑AI tools company, as part of a strategy to offer an end‑to‑end enterprise generative‑AI stack. (forbes.com)
• S&P Global (enterprise data and analytics software) acquired ProntoNLP, a provider of generative‑AI tooling used to derive insights from structured and unstructured data, integrating it into its Market Intelligence software division. (prnewswire.com)
• Enterprise SaaS/platform companies such as Shopify (acquiring AI search startup Vantage Discovery), Advisor360° (acquiring generative‑AI meeting‑notes firm Parrot AI), and Workato (acquiring generative‑AI support‑automation platform DeepConverse) similarly bought AI tooling startups to deepen their AI capabilities for customers. (businessinsider.com)
• Data‑infrastructure provider Oxylabs acquired web‑scraping startup ScrapingBee while building out AI‑powered tools like OxyCopilot and AI Studio, further illustrating infra companies buying AI‑enhanced tooling. (en.wikipedia.org)

These and other deals in 2024–2025 are directly in line with Friedberg’s claim that companies similar to Databricks and Snowflake would need to “strap on tools like this” via further AI/tooling M&A, so the prediction has come true.

Evidence since July 2023 matches Sacks’ prediction of continued AI tooling/infrastructure M&A at high valuations driven by strategic value to large infra providers:

• Databricks → Tabular (data lakehouse infra, June 2024): Databricks agreed to buy data-management startup Tabular to strengthen its open data-lakehouse and AI stack. Reports put the price at over $1 billion, with later reporting saying Databricks "reportedly paid nearly $2 billion" while Tabular had only about $1 million in ARR, implying an extremely high revenue multiple. The deal was explicitly framed as part of the Databricks–Snowflake battle over data formats and AI infrastructure, i.e., strategic toolchain assembly, not current revenue. (techcrunch.com)

• OpenAI → Rockset (vector search / real‑time analytics infra, June 2024): OpenAI acquired Rockset, a real‑time search and database analytics company whose tech (vector search, real‑time indexing) is being integrated directly into OpenAI’s infrastructure to improve ChatGPT and enterprise AI capabilities. Rockset had raised about $105M and was an infrastructure/tooling company rather than a large revenue business, and OpenAI described the deal in clear strategic terms (strengthening its platform) rather than as a revenue play. (reuters.com)

• OpenAI → Windsurf and io (2024–2025): Reports describe OpenAI’s acquisition of AI coding tool Windsurf for around $3 billion and hardware/AI-device startup io (Jony Ive’s company) for about $6.4–6.5 billion, despite io not yet having shipped products. These are framed as large, strategic acquisitions to deepen OpenAI’s end‑to‑end AI product stack (coding tools plus hardware/device layer), again reflecting high valuations driven by strategic positioning rather than current revenue. (timesofindia.indiatimes.com)

• Snowflake → Crunchy Data (Postgres for AI agents, 2025): Snowflake agreed to acquire PostgreSQL specialist Crunchy Data for about $250 million, explicitly to enhance Snowflake’s ability to help customers build AI agents on their data. This follows Snowflake’s earlier acquisition of AI-search startup Neeva and mirrors Databricks’ acquisitions, underscoring that major data/infra platforms are buying key components to complete their AI toolchains. (wsj.com)

• Pattern of high multiples and strategic positioning: Earlier landmark deals like Databricks’ 2023 acquisition of MosaicML (announced just before the podcast but closed after) at $1.3B on roughly $20M revenue (~65× multiple) were already seen as unusually rich and explicitly justified as strategic for generative‑AI infrastructure. Subsequent deals (e.g., Tabular’s billion‑plus valuation on roughly $1M ARR) continued this pattern of large infra providers overpaying on revenue multiples to secure critical AI tooling components. (cbinsights.com)

Overall, from mid‑2023 through 2025 there has indeed been substantial, ongoing M&A in AI tooling and infrastructure, led by major data/AI platforms (Databricks, Snowflake, OpenAI) buying strategic components at very high valuation multiples, exactly in line with Sacks’ reasoning. Hence the prediction is best classified as right.

Sacks argued in July 2023 that an already “huge crunch in late‑stage financing” would only get worse over the next 18 months.

Evidence shows a sharp worsening through late 2023, but then a meaningful rebound during 2024, especially for late‑stage and AI‑driven companies:

• By mid‑2023, the late‑stage market was already in crisis: TechCrunch reported that late‑stage/Series D+ valuations had collapsed (about 60% year‑over‑year and 33% quarter‑over‑quarter in Q2 2023) and that late‑stage deal counts were falling sharply, with late stage hit hardest in the downturn. PitchBook likewise noted that in 2023 the late stages saw valuation multiples and capital supply dwindle, and that Series D+ companies were waiting a record ~1.8 years between rounds. CB Insights’ 2023 report showed global late‑stage median deal size down more than 50% from 2021, with investors shying away from large late‑stage rounds.

  (Sources: TechCrunch on the late‑stage market in August 2023; PitchBook’s US VC market commentary on 2023; CB Insights State of Venture 2023.)

• Conditions did worsen into late 2023: Q4 2023 was described by CB Insights as the harshest quarter for global VC in six years, with funding and deal volume at multi‑year lows and late‑stage mega‑rounds especially scarce. Crunchbase similarly reported that Q4 2023 was the weakest quarter of the year in North America, with late‑stage dealmaking particularly muted.

  (Sources: CB Insights State of Venture 2023; Crunchbase year‑end 2023 funding review.)

• However, by 2024 the market began to recover rather than “only get worse.” PitchBook/NVCA’s data for 2024 show late‑stage deal value rising around 10% quarter‑over‑quarter early in 2024 and a “slight rebound” in late‑stage deals by Q4 2024, with median deal sizes trending upward. A TechCrunch analysis of PitchBook data reported that by the first half of 2024, median valuations for new early‑ and late‑stage deals in the U.S. had not only recovered from 2023 lows but reached all‑time highs. CB Insights’ 2024 report shows Q4 2024 global funding at a two‑year high, with 60% of that quarter’s capital coming from $100M+ mega‑rounds—mostly late‑stage deals. Reuters and other summaries of the 2024 PitchBook‑NVCA Venture Monitor note that total VC investment (especially in AI) grew significantly in 2024 vs. 2023, even though activity remained below 2021 levels and still skewed toward a few large winners.

  (Sources: PitchBook/NVCA Venture Monitor 2024 commentary; TechCrunch on 2024 valuation rebound; CB Insights State of Venture 2024; Reuters on 2024 U.S. VC and AI funding.)

• The environment in 2024 was still challenging—fundraising by VC funds was down, exits were constrained, and capital was highly concentrated in top firms and AI leaders—so a crunch persisted, particularly for non‑elite late‑stage companies. But key indicators like deal value, late‑stage mega‑round activity, and median late‑stage valuations improved vs. the 2023 trough and even vs. mid‑2023, contradicting the idea that financing conditions would only deteriorate throughout the entire 18‑month window.

Because late‑stage financing conditions first worsened but then clearly rebounded within the forecast horizon, Sacks’s directional claim that it was "only" going to get worse over the next 18 months does not hold. The prediction is therefore wrong.

Chamath’s prediction is explicitly about ultimate outcomes for the ~1,400 unicorn startups existing around mid‑2023: what fraction will eventually go to zero, how many will just return capital, and the final blended multiple on invested capital (MOIC) once all is realized. That requires waiting until most of that cohort has either exited (IPO/M&A), been written down to near‑zero, or liquidated.

As of late 2025, that has clearly not happened yet. Crunchbase data shows a persistent "unicorn backlog": its Unicorn Board is close to 1,600 active unicorns, with only around 40 having dropped off the list through closure or down‑rounds since 2022, meaning the vast majority remain private and unliquidated. (news.crunchbase.com) TechCrunch notes that in the most recent decade only about 7% of unicorns have exited, versus 66% in the prior decade, illustrating how few outcomes have actually been crystallized. (techcrunch.com) Other analyses likewise emphasize that exits are scarce and the unicorn ranks remain swollen, not yet worked through via IPOs, acquisitions, or failures. (news.crunchbase.com)

Because most of the mid‑2023 unicorn cohort is still alive on paper, with slow exit markets and limited formal write‑offs, we do not yet observe whether ~60% will “go to zero,” how many will merely return capital, or whether the final blended MOIC for that cohort will approximate 1.1x. There is no comprehensive dataset today that could credibly fix those eventual portfolio‑level results. For that reason, the prediction’s accuracy cannot be determined yet, even though we have partial signals about valuations and fundraising conditions.