IPOs and SPACs are Back, Mag 7 Showdown, Zuck on Tilt, Apple's Fumble, GENIUS Act passes Senate

The prediction sets a timeframe of about one year from the June 2025 recording, i.e., roughly by mid‑2026. Today is November 30, 2025, so less than a year has passed since the podcast release on June 21, 2025, and we have not yet reached the prediction’s target date. Because the deadline for the forecast (mid‑2026) is still in the future, it’s too early to judge whether the video AI model ("Vo 3" / Google Veo) will have “destroyed Hollywood” as claimed.

Additionally, the notion that Hollywood is “done” is somewhat subjective and would be difficult to measure precisely even once the date arrives, but the primary reason we cannot evaluate this forecast yet is simply that the specified time horizon has not elapsed.

The prediction’s own timeframe is “in the near term,” which we normalized as within a few years from June 2025. As of November 30, 2025, only about five months have passed, so the window for judging it has clearly not expired.

Substantively, there are early signs consistent with the prediction:

• Huawei is mass‑shipping Ascend 910C AI accelerators built on SMIC’s domestic 7 nm process as an alternative to Nvidia’s H100/H20 in China; these chips reach roughly 60% of an H100’s performance and are being deployed in large CloudMatrix systems.(mihutz.com)
• Huawei has announced the Ascend 920 on SMIC’s 6 nm node, aiming to exceed Nvidia’s China‑specific H20 in raw performance, though independent benchmarks and large‑scale deployment are still pending.(tomshardware.com)
• U.S. export controls and recent Chinese regulatory moves limiting new Nvidia deployments inside China are pushing domestic firms toward local AI silicon (e.g., Huawei Ascend), creating the beginnings of a structural, China‑based challenger in that market.(reuters.com)

However, as of late 2025 these developments have not yet translated into a clear, global “significant competitive threat” to Nvidia’s overall AI‑chip dominance:

• Analyses consistently find Huawei’s Ascend line still one generation behind Nvidia in performance per watt and ecosystem maturity; training on Ascend often costs more and carries higher integration risk versus Nvidia’s CUDA‑based stack.(ainvest.com)
• SMIC’s 7 nm and 6 nm processes lag TSMC’s leading‑edge nodes that fabricate Nvidia’s top chips, and Huawei’s 2025 production volumes (a few hundred thousand advanced AI chips) are far below Nvidia’s global scale, limiting immediate competitive impact outside China.(evervolve.com)

Overall, China has started developing advanced AI chips on domestic manufacturing processes that challenge Nvidia within China, but it is too early in the stated multi‑year window to say whether this will mature into the “significant competitive threat” to Nvidia’s global dominance that the prediction describes. Hence the status is inconclusive (too early) rather than clearly right or wrong.

Chamath’s claim is explicitly long‑dated: that “over the coming years” essentially all major enterprise/business software will be rebuilt end‑to‑end using new AI‑enabled development toolchains, displacing the current generation of vertical SaaS and custom software.

As of November 30, 2025:

• AI development tools are being adopted very rapidly. GitHub Copilot and similar tools are now used by a large share of professional developers and enterprises. Gartner forecasts that by 2028, about 90% of enterprise software engineers will use AI code assistants, up from <14% in early 2024.(github.blog) Surveys also report that a majority of professional developers use AI tools regularly, and Copilot alone has tens of thousands of enterprise customers and millions of users.(secondtalent.com) This supports the direction of his thesis (AI deeply permeating the software development lifecycle).

• However, the existing SaaS and enterprise software base is still enormous and largely intact. The global SaaS market in 2025 is hundreds of billions of dollars and is projected to keep growing through 2030–2034, not to be wholesale displaced in the near term.(rss.globenewswire.com) There is no evidence that “essentially all” major enterprise systems have been or are imminently being rebuilt from scratch; instead, vendors are adding AI features, consolidating, or modernizing parts of their stacks rather than replacing everything soup‑to‑nuts.(timesofindia.indiatimes.com)

• Timeframe: we are only ~5 months past the June 21, 2025 episode. A prediction about “the entirety of the software that runs the world” being rebuilt “over the coming years” spans many years, not months. The horizon he implies (multi‑year global rebuild) is far beyond November 2025, so it is too early to say definitively whether this end‑state will or will not materialize.

Because (a) AI toolchain adoption is clearly underway in line with the spirit of the claim, but (b) the outcome he specifies—nearly all major enterprise/business software rebuilt end‑to‑end—concerns a much longer period than has elapsed, the prediction cannot yet be judged as either correct or incorrect.

Therefore the status as of November 30, 2025 is “inconclusive (too early to tell).”

The prediction explicitly sets a long horizon: performance dispersion in the S&P 493 “over roughly the next decade.” As of now (late November 2025), only a few months have passed since the June 21, 2025 podcast release date, far too little time to evaluate whether:

• AI adoption has persistently driven “very large performance dispersion” across the S&P 493, and
• this has in fact produced “some of the largest relative‑return opportunities for public‑equity investors in decades.”

While there are already anecdotal signs of differing AI strategies and some early market enthusiasm around AI-forward companies, that short-term behavior cannot conclusively confirm or refute a decade-scale claim about structural return dispersion and opportunities. We would need many more years of realized returns and clear attribution to AI-driven operational changes to make a fair assessment.

Because the forecast window is still in its very early phase, the correct classification at this time is “inconclusive (too early)”.

The prediction is explicitly about "the coming decade" of performance dispersion within the S&P 493 between AI adopters and laggards. The podcast was released on 21 June 2025, and the current date is 30 November 2025—only a few months later. That is far too short a window to evaluate a decade‑scale claim about structural dispersion in equity returns and the resulting opportunity set for stock pickers. Even if early patterns (e.g., some AI‑exposed names outperforming) can be observed, they cannot yet be reliably attributed to the long‑run effect the prediction describes. Therefore, it is too early to determine whether this forecast is right or wrong.

The prediction explicitly uses a 5–10 year horizon for these newly public companies to compound revenues around 25% annually and become the next generation of "compounders." As of November 30, 2025—about five months after the June 21, 2025 podcast—CoreWeave, Circle, and Chime have only a short public track record, even though early growth looks strong (e.g., CoreWeave’s 2024 revenue grew about 737% year over year and it guides to multi‑billion‑dollar 2025 revenue; Circle reported 53% year‑over‑year revenue growth in Q2 2025, its first quarter as a public company; Chime’s revenue grew roughly 30% in 2024 and continues to rise post‑IPO). (fool.com) Whether many of these firms will sustain ~25% annual growth over an entire 5–10 year period and truly become durable compounders cannot yet be determined, so the prediction is too early to judge.

As of November 30, 2025, there isn’t enough evidence to say that Chamath’s structural, long‑term call on traditional enterprise SaaS has clearly proven right or wrong.

Key points:

• Sector growth has slowed, but not collapsed. Large flagship SaaS vendors like Salesforce are still growing revenue in the high single‑ to low double‑digit range (around 7–10% YoY in recent quarters), even though that’s a clear deceleration from the mid‑teens or higher growth rates seen earlier in the 2020s. (reddit.com) This supports the “growth deceleration” part of his thesis but does not yet show that the business model has broadly failed.

• Investor returns and expectations are mixed. Salesforce, one of the emblematic cloud SaaS names, is down roughly 30% year‑to‑date in late 2025 and has modest returns over five years, underperforming the S&P 500, with forward revenue growth now expected to be <10% annually and valuation multiples compressed. (barrons.com) At the same time, Salesforce still guides to over $60B in revenue by 2030 and is executing large AI‑driven acquisitions like Informatica, signaling that markets do not view the category as “over” so much as repriced. (reuters.com) Cloud‑heavy vendors like Oracle are even guiding to higher cloud growth going forward. (wsj.com) So structural underperformance is not clearly established across the whole space.

• Public cloud/SaaS indexes show pressure but continued viability. The BVP Nasdaq Emerging Cloud Index (a proxy for public cloud/SaaS names) remains well above its 2018 base value (around 1,600–1,700 vs. 1,000 at inception), indicating that the segment is still meaningful. (indexes.nasdaqomx.com) It has had bouts of underperformance and drawdowns—Nasdaq’s March 2025 scorecard highlighted cloud and AI thematic indexes among the worst performers that month, with the cloud index down ~11.4%—but this is consistent with a repricing and rotation, not clear evidence that the “jig is totally up for software.” (nasdaq.com)

• Customers are experiencing tool fatigue and poor ROI. Surveys like Freshworks’ 2025 research show enterprises wasting roughly 20% of software budgets on unused tools, failed implementations, and hidden costs, with significant productivity loss and burnout due to software complexity and tool sprawl. More than half of businesses report that software investments aren’t delivering the expected ROI. (itpro.com) Other analyses quantify overlapping SaaS tools, unused licenses, and integration overhead, and explicitly argue for moving from “SaaS sprawl” to more consolidated, often AI‑driven, custom architectures. (medium.com) This supports the direction of Chamath’s critique (customers are tired of “yet another tool”).

• AI and agentic systems are surging, but mostly as overlays, not wholesale replacements. Enterprise adoption of AI agents and compound/agentic AI architectures is ramping quickly, and many forecasts expect strong ROI and rapid market growth through 2030. (en.wikipedia.org) However, in practice, much of this is happening inside existing SaaS platforms (e.g., Salesforce’s Agentforce and Data Cloud, or CRM‑style vendors adding copilots) rather than via enterprises replacing large swaths of SaaS with entirely custom-built AI software stacks. (marketwatch.com) That pattern is consistent with evolution and integration, not yet with the broad displacement implied by “the jig is totally up for software.”

• Structural claims need a longer horizon. Chamath was explicitly talking about a multi‑year structural shift driven by AI making it cheap to rebuild vertical software and by enterprises rejecting per‑seat/ever‑escalating license models. (metacast.app) Only about five months have passed since the June 21, 2025 episode; even where we see funding pressure and consolidation in enterprise applications (deal funding down modestly, more M&A, lower valuations), that is still early and influenced by macro conditions and the post‑2021 repricing as much as by AI‑driven custom solutions per se. (timesofindia.indiatimes.com) We do not yet have enough multi‑year data on ROE, sector‑wide returns, or actual replacement of SaaS by custom AI to decisively validate or falsify his structural forecast.

Putting these strands together: some elements of his narrative (slower growth, multiple compression, tool fatigue, rising AI/agent adoption) are clearly observable, but the stronger claim—that traditional enterprise SaaS as a whole will structurally underperform and be displaced by cheaper AI‑driven custom software—cannot reasonably be judged after only a few quarters. The available evidence is consistent with both “repricing and adaptation” and “early stages of eventual disruption,” so the prediction’s ultimate accuracy remains inconclusive at this time.

Only about five months have passed between the podcast release (21 June 2025) and the current date (30 November 2025), and the claim is explicitly framed as “over the long term”. That horizon is typically several years, so there hasn’t been enough time to decisively confirm or falsify the prediction.

Current evidence actually points the opposite way in the short run:

• Snowflake continues to center its business on consumption‑based pricing and is actively promoting guides and playbooks that advocate usage‑based models, not retreating from them. (snowflake.com)
• Independent writeups of Snowflake’s pricing in 2025 still describe the familiar pay‑for‑what‑you‑use model across compute and storage (credits per warehouse size, TB‑per‑month storage), rather than a shift to a fundamentally different structure. (blog.twingdata.com)
• Specific tweaks, like Snowpipe moving to a fixed credit amount per GB for ingestion, simplify billing but remain per‑usage; they don’t abandon the consumption framework. (docs.snowflake.com)
• Financially, Snowflake’s product revenue is still growing >20% annually with strong net‑revenue‑retention and repeated guidance beats in 2025, which suggests the model is economically viable so far, even if customers are more cost‑conscious and optimize usage. (reuters.com)
• There is no clear evidence of large‑scale customer migration from Snowflake to lower‑cost Postgres/Supabase‑style alternatives; if anything, we see Snowflake acquiring Crunchy Data to offer its own Postgres option and case studies of firms migrating to Snowflake from other platforms. (constellationr.com)

Given the explicitly long‑term nature of the prediction and the relatively short time elapsed, plus the lack of clear evidence either of mass migrations away from Snowflake or of its pricing model collapsing, the fairest assessment as of November 2025 is that the prediction’s truth value is still undetermined.

The prediction explicitly covers “the next five years” from June 21, 2025, i.e., roughly through mid‑2030. As of today (November 30, 2025), less than a year of that five‑year window has elapsed.

Even if we could already observe some dispersion between AI winners and laggards in the Mag 7 or the broader S&P 500, that would not be enough to validate the claim that:

1. This entire five‑year period is unusually good for active stock pickers, and
2. The key driver of that persistent dispersion is AI adoption.

Both points require multi‑year realized performance data (active manager excess returns, cross‑sectional dispersion, sector/industry attribution to AI themes, etc.) that simply do not exist yet for the full horizon. Because the evaluation date (mid‑2030) has not arrived, the prediction cannot be judged right or wrong at this time.

The prediction is framed on a decade-long horizon ("within roughly the next decade"). As of November 30, 2025, only about a year and a half has passed since the June 21, 2025 podcast release.

While there is active discussion and early evidence that AI may pressure some traditional IT services and outsourcing models, we cannot yet determine:

• whether “many” such rolled‑up IT services businesses will, in aggregate, actually lack terminal buyers, and
• whether exit values for private‑equity roll‑ups in those sectors will be structurally and durably depressed specifically because AI agents automate their value proposition.

Both of those require several years of M&A, IPO, and secondary transaction data across the sector, which simply do not exist yet for the 10‑year window implied. Therefore, the prediction’s outcome cannot be assessed this early, even directionally, beyond noting it as a plausible but unproven long‑term thesis.

As of November 30, 2025, only about five months of the 1–2 year window (June 21, 2025 through at least June 21, 2026, and possibly June 21, 2027) have elapsed, so the full period the prediction covers has not yet played out.

Evidence so far is supportive of Friedberg’s view:

• High‑quality tech IPOs have occurred and been very well received. Figma went public on July 31, 2025 at $33 per share and raised roughly $1.2 billion; the stock closed its first day over $115, more than 250% above the IPO price, signaling extremely strong demand for a marquee growth software name. (forbes.com)
• Other “fantastic” growth tech/fintech/AI names have also listed with big pops. Circle Internet Group’s June 2025 IPO and CoreWeave’s March 2025 IPO are cited alongside Figma as standout tech offerings; Circle’s first‑day gain was about 165%, while CoreWeave and other AI‑infrastructure/tech IPOs have delivered triple‑digit returns post‑listing, indicating robust investor appetite for top‑tier growth tech issuers. (finance.yahoo.com)
• Broader US IPO data suggests the market is ‘open for business’ in 2025, especially for quality names. By mid‑2025, U.S. IPOs were averaging first‑day pops of around 27–36%, near decade highs, with tech leading the way and 241 U.S. IPOs year‑to‑date—evidence of an active, receptive market. A February 2025 Renaissance Capital review also noted that most sizable IPOs were trading above offer, even if a few large deals dragged down weighted averages. (finance.yahoo.com)

However, Friedberg’s prediction is explicitly about the market remaining open over the next 1–2 years. Conditions could still change (e.g., macro downturn, rate shock, regulatory or geopolitical events) between now and mid‑2026/2027 that might shut the window for high‑quality growth tech IPOs. Because that future part of the timeframe has not yet occurred, we cannot definitively say the prediction has proven right or wrong.

Therefore the correct status at this time is inconclusive (too early): early developments are consistent with his thesis, but the full 1–2 year horizon has not yet passed.

The prediction is explicitly about “the coming years” of the AI transition, implying a multi‑year horizon for evaluating whether a long/short strategy (short S&P 500, long a few emerging AI “category killers”) outperforms simple passive S&P 500 exposure.

The episode containing this remark was released on June 21, 2025.(tapesearch.com) As of today, November 30, 2025, less than one year has passed, which is well short of a reasonable interpretation of “years” (at least ~2–3 years). In addition, the prediction does not specify:

• which exact AI “category killer” stocks to hold,
• the position sizing and rebalancing rules,
• or the precise evaluation window.

Because (1) the stated timeframe has not elapsed, and (2) the strategy is too loosely defined to test rigorously so far, there is not enough information or elapsed time to determine whether Friedberg’s claim about this strategy being unusually attractive and likely to outperform the S&P 500 has proven correct or incorrect.

So the status as of November 30, 2025 is “inconclusive (too early)”.

It’s too early to evaluate this prediction.

• The prediction horizon is “roughly five years from now” relative to the podcast date of June 21, 2025, i.e., around mid‑2030.
• Today is November 30, 2025, only a few months after the prediction, so we don’t yet know Microsoft’s global employee count in 2030.
• Current public data only shows Microsoft’s workforce up to recent fiscal years (roughly in the 220,000–250,000 range, depending on year and layoffs), but nothing about the 2030 level yet.

Because the specified five‑year period has not elapsed, the prediction cannot yet be judged; hence the status is inconclusive (too early).

The prediction window is about five years from the podcast date of June 21, 2025, so it runs roughly until mid‑2030. As of November 30, 2025, we are only a few months into that window, so it is too early to judge whether Microsoft’s eventual total headcount will end up below its recent peak (~250,000 employees).

Recent public data suggests Microsoft’s headcount has been near or above that quarter‑million level in the mid‑2020s, but the prediction concerns where it will settle by 2030, not where it is in 2025. Because that future outcome cannot yet be observed, the correctness of the prediction cannot be determined at this time.

Therefore, the status of the prediction is inconclusive (too early).

The prediction is explicitly about “the next several years (on the order of five years)” after the podcast release on 21 June 2025. That implies a horizon out to roughly mid‑2030.

As of today (30 November 2025), only about 5 months have elapsed—far short of the multi‑year timeframe. While we can inspect early data (e.g., Microsoft’s FY 2025 results and any recent headcount or layoff announcements) to see directionally what’s happening, those partial trends cannot reliably confirm or falsify a five‑year operational prediction.

Because the specified time window has not yet passed, and nothing in the current data makes the prediction logically impossible to achieve by 2030, the only defensible assessment is that it is too early to know whether Jason’s prediction about strong revenue growth with roughly flat headcount (~250k ±10%)—and thus much higher revenue per employee—will ultimately be right or wrong.

The prediction specifies a five‑year holding period starting from around the podcast date (June 21, 2025), implying an evaluation point near June 21, 2030. As of the current date (November 30, 2025), less than one year of that five‑year horizon has elapsed. Over such a short fraction of the intended period, any outperformance or underperformance of an equal‑weighted basket of AWS (if it were separable), Microsoft Azure, and Google Cloud Platform versus the broad market would not be a reliable indicator of the long‑term outcome. Because the prediction’s success condition is explicitly about returns over the full five years and we are far from that endpoint, it cannot yet be determined whether this basket will ultimately outperform the broad market to the degree implied by ‘you wouldn’t need to own anything else.’

The prediction specifies a five‑year horizon starting from around the podcast release date of June 21, 2025. That window runs approximately until June 21, 2030. As of today (November 30, 2025), only a little over five months of that period have elapsed, far short of the full five‑year horizon required to judge whether a portfolio concentrated in AWS, Azure (Microsoft), and GCP (Google) has outperformed the rest of the market. Because the prediction is explicitly long‑dated and we are still early in the evaluation period, it cannot yet be determined whether it is right or wrong.

The GENIUS Act is the federal stablecoin bill discussed on the podcast. After the Senate passed it on June 17, 2025, the House of Representatives passed the bill on July 17, 2025, by a vote of 308–122. President Donald Trump signed it into law on July 18, 2025, making it Public Law 119‑27.

These dates are documented in:

• The GENIUS Act legislative history, showing House passage on July 17, 2025, and Trump’s signature on July 18, 2025. (en.wikipedia.org)
• Congress.gov, which lists S.1582 (GENIUS Act) as having "Became Public Law No: 119‑27" on July 18, 2025. (congress.gov)
• Contemporary news coverage confirming Trump signed the GENIUS Act on July 18, 2025. (politico.com)

The podcast aired June 21, 2025. The House acted 26 days later (about three and a half weeks), which reasonably fits the colloquial "next few weeks," and Trump did in fact receive and sign the bill during his current term. Both key elements of Sacks’s prediction—timing in the near future and ultimate presidential signature—were fulfilled, so the prediction is best judged right.

Key facts:

• The GENIUS Act, a federal framework for U.S. dollar stablecoins, was passed by Congress and signed into law on July 18, 2025, creating a new regulatory category for payment stablecoins that are not bank deposits (they are explicitly treated separately from traditional bank deposits and securities/commodities). (en.wikipedia.org)

• Since then, no major U.S. bank has launched a retail or broadly-available GENIUS‑Act "payment stablecoin":

  • Bank of America repeatedly says it is ready and expects to launch a fully dollar‑backed stablecoin but only once the regulatory path is fully clear; as of mid‑ to late‑2025 it still describes this as a future plan, with no product or launch date announced. (coindesk.com)
  • Citigroup likewise says it is considering or looking at the issuance of a Citi stablecoin for digital payments, but again only as an exploration, not a launched product. (reuters.com)
  • An August 12, 2025 Reuters piece notes that after the GENIUS Act, firms like Bank of America and Citigroup are exploring GENIUS‑compliant stablecoins, but stresses that deployment strategies, regulatory details, and timelines remain unresolved. (reuters.com)

• The one concrete on‑chain dollar instrument from a major U.S. bank is JPMorgan’s JPMD:

  • JPMorgan piloted and then, on November 12, 2025, made its USD deposit token "JPMD" available to institutional clients on Coinbase’s Base L2. (jpmorgan.com)
  • JPMorgan itself and multiple commentators stress that JPMD is a deposit token, not a GENIUS‑Act payment stablecoin: it represents insured bank deposits, accrues interest, and is framed as "commercial bank money" on‑chain. JPMorgan’s Kinexys unit explicitly says, “This is a deposit token and not a stablecoin; clients are sending and receiving real bank money on Base,” and the Financial Times similarly contrasts JPMD with stablecoins for regulatory purposes. (en.coinotag.com)

Interpretation:

• Sacks’ prediction was that after the GENIUS Act, major U.S. banks would issue their own U.S.-dollar stablecoins under the new framework.
• As of November 30, 2025:
  • Banks are clearly preparing and signaling intent to issue such stablecoins under GENIUS. (coindesk.com)
  • But no major U.S. bank has yet launched a GENIUS‑regulated payment stablecoin; JPMorgan’s live product is deliberately structured and branded as a deposit token rather than a GENIUS stablecoin, and other big banks remain in the planning phase.

Because the GENIUS Act only became law in July 2025 and large banks are still in the design/regulatory‑approval stage, it is too early to say that the prediction—"the banks ... are going to issue one" under this new framework—has failed. At the same time, the exact outcome Sacks described has not yet materialized.

Conclusion: The prediction is neither clearly right nor clearly wrong as of November 30, 2025; it remains inconclusive (too early).

The GENIUS Act was signed into U.S. law on July 18, 2025, and it includes a three‑year transition period after enactment before U.S. digital asset service providers are generally barred from offering payment stablecoins that are not issued by a "permitted payment stablecoin issuer"; foreign issuers can continue if they meet certain conditions to comply with U.S. lawful orders and reciprocal arrangements, rather than being strictly forced to redomicile onshore. 【1view0【【2view0【【3view0【 That three‑year window runs until July 18, 2028, which is still in the future as of November 30, 2025, so the time horizon of the prediction has not elapsed. 【1view0【【0news16】

Empirically, since passage of the GENIUS Act, Tether has announced plans to: (a) use the law’s foreign‑issuer pathway to make USDT available in the United States while remaining an offshore issuer, and (b) launch a separate, U.S.-based compliant stablecoin (variously reported as USAT/USA₮) issued through Anchorage Digital Bank under U.S. regulation. 【4search0【【4search8【【4news12【【4news13【 These steps do not yet amount to fully relocating Tether’s existing offshore stablecoin operations onshore, but because the statutory three‑year deadline has not arrived, it is still possible (though not guaranteed) that Tether and similar issuers could restructure to satisfy the prediction by 2028. Given the open time window and ongoing strategic moves, the prediction cannot currently be judged true or false.

The GENIUS Act is a U.S. stablecoin-regulation bill that passed the Senate on June 17, 2025, then went to the House.

Official legislative histories and news coverage show that the U.S. House of Representatives passed the GENIUS Act on July 17, 2025, and President Trump signed it into law on July 18, 2025. (en.wikipedia.org)

Since the podcast episode was released on June 21, 2025, the bill cleared the House less than a month later—well within a reasonable interpretation of “very quickly” in congressional terms. Therefore, Jason’s prediction that the referenced crypto bill (the GENIUS Act) would pass the House in the near term was borne out by events.