Elon gets paid, Apple's AI pop, OpenAI revenue rip, Macro debate & Inside Trump Fundraiser

What happened: In the 2024 U.S. presidential election on November 5, 2024, Donald Trump (Republican) ran against Kamala Harris (Democrat), not Joe Biden. Trump won with 312 electoral votes to Harris’s 226, and about 49.8% of the popular vote to Harris’s 48.3%.(en.wikipedia.org)

On “Democrats will lose”: This part is correct in a narrow sense: the Democratic ticket (Harris/Walz) lost the presidency to Trump/Vance.(en.wikipedia.org)

On “Joe Biden will be defeated”: This is wrong. Biden did not end up as the Democratic nominee; Harris was. Since Biden never faced Trump on the 2024 ballot, he could not be “demolished” or defeated in that election.(en.wikipedia.org)

On “by a large margin / demolished”: Trump’s national popular-vote margin over Harris was about 1.5 percentage points, the narrowest popular-vote gap since 2000, and the decisive states collectively turned on roughly 230,000 votes across Michigan, Pennsylvania, and Wisconsin—evidence of a close race, not a blowout.(cookpolitical.com) While Trump’s 312–226 Electoral College margin is larger than in some recent elections, major analyses characterize his popular-vote win as modest and historically tight.(cookpolitical.com)

Because the prediction specifically asserted that Joe Biden himself would be the candidate and would get “demolished” (a large-margin defeat), and neither of those key elements occurred—even though Democrats did lose overall—the prediction is best scored as wrong.

Evidence from multiple sources shows that before the November 5, 2024 U.S. presidential election, the Democratic Party did in fact replace Joe Biden with another nominee:

• Joe Biden became the Democratic Party’s presumptive 2024 nominee on March 12, 2024 after clinching enough primary delegates. (en.wikipedia.org)
• Following intense concerns after the June 27, 2024 debate, Biden announced on July 21, 2024 that he was withdrawing from the 2024 presidential race and would not accept the nomination, and he endorsed Vice President Kamala Harris as his preferred successor. (en.wikipedia.org)
• Harris immediately launched her own campaign and, by July 22, 2024, had secured enough delegate support to become the presumptive Democratic nominee in Biden’s place. (en.wikipedia.org)
• On August 5, 2024, in a virtual roll call of Democratic National Convention delegates, Kamala Harris was formally confirmed as the Democratic Party’s 2024 presidential nominee—months before the November 5, 2024 election. (en.wikipedia.org)

So, in the run-up to the 2024 election, Democrats effectively conducted the “hot swap” Jason described: they replaced Biden with another candidate (Harris) as their presidential nominee prior to Election Day. The prediction therefore came true.

Multiple reports show that President Joe Biden did voluntarily end his 2024 reelection bid and step down as the Democratic candidate months before Election Day. On July 21, 2024, Biden publicly announced in a letter that he would “stand down” from the race and focus on his duties as president, and he simultaneously endorsed Vice President Kamala Harris to be the Democratic nominee. (theguardian.com) Coverage from PBS, CBS, and others confirms that this was a voluntary decision; legally, because Biden had already secured nearly all the Democratic delegates, he could not be forced out and therefore had to step aside of his own accord, which he did. (cbsnews.com) Harris subsequently became the official Democratic presidential nominee in August 2024. (en.wikipedia.org) Since Biden both voluntarily withdrew and did not remain the nominee through Election Day 2024, Chamath’s prediction that “I don’t think Biden is going to step down at all” is wrong.

As of November 30, 2025, it is not yet known whether Joe Biden will serve the entire presidential term that would end in January 2029, because that date is still more than three years in the future. Chamath’s prediction is conditional (“if Biden wins, he won’t make it four years”) and can only be definitively evaluated after we observe whether a full term ending in January 2029 is completed. Since that has not yet occurred (or failed to occur), the correctness of the prediction is too early to determine, so it is best classified as inconclusive at this time.

Two separate issues make this prediction impossible to score as right or wrong at this time:

1. The conditional scenario did not fully occur. Joe Biden initially ran for re‑election, but he withdrew from the 2024 race on July 21, 2024, after securing the primary delegates, and endorsed Kamala Harris, who then became the Democratic nominee. Donald Trump became the Republican nominee and went on to win the 2024 election, being inaugurated on January 20, 2025. (en.wikipedia.org)
   The prediction was explicitly conditional on both being candidates in 2024; in the realized timeline Biden did not remain the Democratic candidate into the general election, so the comparison Chamath was describing is now partly counterfactual.

2. The forecast horizon (a full 2025–2029 term) has not elapsed. As of November 30, 2025, Donald Trump is serving as president and is publicly active (e.g., pursuing his policy agenda and even musing about 2028), which indicates he is presently functioning in office. (theguardian.com) However, the claim was about which man had a higher probability of remaining physically and cognitively capable of serving out an entire four‑year term starting in January 2025. Trump’s current term runs until January 20, 2029, so whether he will remain fully capable for that entire period is not yet knowable.

Because (a) the condition about both being candidates did not hold in the way implied, and (b) the full four‑year period has not yet passed, the prediction cannot currently be evaluated as correct or incorrect, so it is inconclusive (too early to tell).

Available reporting shows that Apple has not auctioned off its system‑level LLM integration to the highest bidder, nor set up a search‑style paid competition for default placement.

1. OpenAI / ChatGPT integration: When Apple announced Apple Intelligence and system‑wide ChatGPT integration at WWDC 2024, multiple reports (summarizing Bloomberg’s Mark Gurman) stated that neither Apple nor OpenAI is paying the other for the deal, and that the arrangement is not expected to generate meaningful revenue initially. Instead, Apple grants OpenAI exposure to hundreds of millions of devices, while Apple gets ChatGPT features for free. This is explicitly contrasted with the lucrative Google‑as‑default‑search deal and described as not involving direct payments. (gadgets360.com) Apple Intelligence documentation likewise describes ChatGPT as an opt‑in, system‑wide integration, free to users, with optional sign‑in for paid ChatGPT features, and notes Apple plans to support other models in the future. (en.wikipedia.org) None of this resembles an auction for default status.

2. Future revenue‑sharing, not auctions: Follow‑on coverage notes Apple may in the future receive a cut from third‑party LLM vendors who monetize on Apple platforms (e.g., ChatGPT Plus subscriptions via in‑app purchase), and that it is in talks with other LLM providers such as Google and Anthropic. But these are described as potential revenue‑sharing or partnership models, not as a bidding war for default placement. (thedailyjagran.com) No source indicates Apple has run or announced a formal auction for the system‑level LLM slot.

3. Google Gemini / Siri deal structure: In late 2025, Bloomberg/Reuters reporting indicates Apple is paying Google roughly $1 billion per year for a custom Gemini model to power a redesigned Siri as part of Apple Intelligence, after evaluating OpenAI and Anthropic. (reuters.com) That is the opposite direction of the classic Google‑Search‑on‑iPhone arrangement (where Google pays Apple); it reflects Apple choosing a vendor and paying for access, not running a highest‑bidder auction in which model providers pay Apple for default placement.

Across the 2024–2025 coverage, there is no evidence that Apple has auctioned off LLM integration rights or default status to the highest bidder, or that OpenAI, Google, or others have paid Apple for that privilege. Instead we see (a) a non‑monetary, opt‑in OpenAI integration and (b) Apple itself paying Google for Gemini. That contradicts the specific prediction that Apple would monetize its default LLM slot via a paid auction similar to its Google search deal.

Jason’s prediction is about the next AI adoption cycle, i.e., a multi‑year future trajectory, not a specific state by late 2025. That cycle is clearly still unfolding, so it’s too early to say definitively whether Apple will ultimately become the dominant consumer AI assistant platform.

As of November 30, 2025, Apple has launched Apple Intelligence across iOS, iPadOS, macOS and VisionOS, but rollout is still constrained by hardware (A17‑class and newer) and staged language/region support, indicating an early and incomplete deployment rather than a mature, settled market position. (de.wikipedia.org) Apple has also moved to license Google’s Gemini model for a redesigned Siri, with major Siri upgrades not expected until 2026, which shows Apple is still catching up technically rather than already having secured a clear, enduring lead. (reuters.com)

Across personal devices, Apple’s installed base remains much smaller than competitors on key platforms where AI assistants run: Android holds about 72% of the global mobile OS market versus ~28% for iOS, and Windows dominates desktop/laptop OS share, which underpins broad deployment of Microsoft’s Copilot and other assistants. (en.wikipedia.org) In adjacent consumer‑assistant arenas like smart speakers and smart‑home voice AI, Amazon Alexa and Google Assistant still lead, with Apple Siri/HomePod participating but not in a clearly dominant position. (market.us)

Taken together, current evidence shows Apple is a strong but not dominant player in consumer AI assistants as of late 2025, while the broader “AI adoption cycle” for assistants on phones, tablets, and PCs is still in its early to middle stages. Because the prediction is explicitly about who will win over that entire cycle, and the outcome has not yet been determined, the correct status today is inconclusive (too early) rather than right or wrong.

1. What actually happened in the Tesla fee case

• The plaintiffs’ lawyers in Tornetta v. Musk asked for a fee in Tesla stock worth well over $5–7 billion.
• In March 2024, Delaware Chancellor Kathaleen McCormick rejected that multi‑billion‑dollar request and instead awarded about $345 million in fees (cash or stock) — a huge sum, but not “billions of dollars” as Sacks’ conditional referred to.
  • Associated Press and Bloomberg coverage of the decision report the fee at roughly $345 million, far below the multi‑billion request. (apnews.com)

So the explicit trigger in the normalized prediction — Delaware awarding the lawyers billions of dollars in fees — did not occur.

2. What happened to Delaware’s status as a corporate hub

Since the January 2024 ruling voiding Musk’s Tesla pay package, there has been a noticeable but still modest “DExit” trend:

• A Harvard Law Forum empirical study finds that from 2024 through mid‑2025, Delaware lost a net 11 large public companies (market cap > $250m) to reincorporations elsewhere (16 left, 5 moved in). (corpgov.law.harvard.edu)
• The same analysis notes Delaware still accounts for ~80–90% of IPO incorporations, with only a slight dip in its share in early 2025. (thegamingboardroom.com)
• A broader overview of the “Delaware corporate exodus” reports at least a dozen billion‑dollar public companies reincorporating away by late 2025, and an 8% drop in new business formations, but emphasizes there has been “no mass exodus”; defections are high‑profile but numerically small. (en.wikipedia.org)
• Delaware still hosts roughly two‑thirds of Fortune 500 companies and more than 2.1–2.2 million registered entities, and remains widely described by officials and legal experts as the premier corporate domicile, albeit under new competitive pressure from Texas and Nevada. (en.wikipedia.org)

So there is some material movement away from Delaware, but Delaware has not clearly lost its status as the leading U.S. state for incorporation; it is still dominant while facing increased competition and modest erosion at the margin.

3. Why the forecast is scored as ambiguous

The normalized prediction is explicitly conditional:

If the Delaware court awards the plaintiff’s lawyers billions in fees, then Delaware will suffer a significant loss of premier status, with a material shift of incorporations elsewhere.

• The antecedent (billions awarded) never happened; the court capped the fee at ~$345m, not billions. (apnews.com)
• Nonetheless, some exodus did occur after the Tesla ruling and broader Chancery decisions, but:
  • It is modest relative to Delaware’s enormous base, and Delaware clearly remains the primary incorporation venue.
  • The observed shifts are driven by a mix of factors (judicial scrutiny of founders, overall legal climate, tax and cost considerations), not specifically by a multi‑billion fee award that never materialized. (en.wikipedia.org)

Because the key condition in Sacks’ forecast (a multi‑billion‑dollar fee award) did not occur, we cannot directly test his stated “if X then Y” claim. At the same time, Delaware’s dominance has been dented but not clearly “lost”, making it impossible to give a clean “right” or “wrong” verdict on the normalized consequence even by analogy.

Given those factors, the fairest evaluation is “ambiguous”: the conditional was never triggered, and the real‑world outcomes (modest DExit, continued Delaware dominance) do not let us definitively confirm or falsify the full, conditional prediction.

All three concrete elements of Chamath’s near‑term macro call have already occurred within the 2024–2025 rate‑setting cycle:

1. Unemployment rose from mid‑2024 levels.

   • In June 2024 (around the time of the podcast), the U.S. unemployment rate was 4.1% (seasonally adjusted). (bls.gov)
   • By August and September 2025, it had climbed to 4.3%–4.4%, the highest since 2021, indicating a clear rise from that 4.1% baseline. (tradingeconomics.com)

2. At least one quarter of negative real GDP growth occurred.

   • BEA data show real U.S. GDP decreased in Q1 2025, with the third estimate recording a ‑0.5% annualized change from Q4 2024 (after Q4 2024 had grown at about 2.4%). (bea.gov)
   • This satisfies his criterion of “at least one quarter of negative real GDP growth” in the 2024–2025 window.

3. The Fed cut rates more than once in the cycle.

   • The FOMC reduced the federal funds target range by a cumulative 100 bps over its September, November, and December 2024 meetings, from 5.25–5.50% down to 4.25–4.50%. That is already three distinct cuts. (federalreserve.gov)
   • The Fed has since continued easing, including another 25 bp cut in October 2025 to a 3.75–4.00% range, confirming multiple further cuts within the same rate‑setting cycle. (federalreserve.gov)

Given that unemployment is higher than at the time of the prediction, real GDP has logged a contracting quarter, and the Fed has executed several rate cuts within 2024–2025, Chamath’s composite prediction is substantively correct within the timeframe he specified.

BEA data show that all U.S. real GDP prints leading into the November 5, 2024 election were positive:

• Q1 2024: Real GDP grew at a 1.4% annual rate (third estimate). (bea.gov)
• Q2 2024: Real GDP grew at a 3.0% annual rate (second estimate). (bea.gov)
• Q3 2024: Real GDP grew at a 2.8% annual rate (advance and second estimates). (bea.gov)

Other compiled series of quarterly real GDP also show continuous positive growth from Q1–Q4 2024, with Q3 and Q4 2024 at 3.1% and ~2.4% respectively, and the first negative quarter only appearing in Q1 2025 (about –0.3%). (visualcapitalist.com)

Chamath’s claim was that it was “very possible” (a material chance) that there would be a negative, recessionary GDP print going into the election. In reality, the quarterly GDP growth rate immediately preceding the election (Q3 2024) was clearly positive, and no negative GDP quarter occurred before November 5, 2024. Thus, judged on the event outcome (whether such a negative pre‑election print actually happened), the prediction did not come true.

Contemporaneous U.S. data for mid‑2024 do not support classifying the macro environment as stagflation under the usual definition of high inflation plus stagnant/low growth and typically elevated unemployment.(en.wikipedia.org)

Growth: According to the BEA, real GDP grew at a 1.4% annual rate in Q1 2024 and then accelerated to a 3.0% annual rate in Q2 2024 (second estimate).(bea.gov)(bea.gov) That Q2 pace is solidly above most estimates of U.S. trend growth and not consistent with “stagnant” or “1–2%” growth as in the normalized prediction.

Inflation: CPI inflation was above the Fed’s 2% target but moderating: the 12‑month CPI increase was 3.3% in May 2024 and 3.0% in June 2024.(fraser.stlouisfed.org)(fraser.stlouisfed.org) The Fed’s preferred PCE measure showed Q2 2024 inflation around the mid‑2% range, with core PCE about 2.8–2.9% annualized.(bea.gov)(bea.gov) That is mildly elevated but not the kind of persistently high inflation usually associated with stagflation.

Unemployment: The unemployment rate in June 2024 was 4.1%, up from the lows of 2022–23 but still historically low and far from the high‑unemployment environment typically seen in stagflation episodes.(cnbc.com)

Putting this together, mid‑2024 featured moderate, above‑trend growth, easing but still‑above‑target inflation, and relatively low unemployment, which does not meet standard criteria for stagflation. The more specific normalized prediction that real GDP growth would remain in the 1–2% range through Q2 2024 is also contradicted by the actual ~3% growth reading. Therefore, Friedberg’s claim that “we have stagflation right now” is best judged as wrong.

The prediction explicitly covers “the next couple of years” after June 2024, i.e., roughly through mid‑2026. As of now (late November 2025) that window has not finished, so the forecast period is still in progress.

Available data are also mixed relative to the normalized claim:

• Headcount levels: Global employment is still growing rather than flat. The ILO’s World Employment and Social Outlook updates project tens of millions of net new jobs in 2024–2025 and forecast employment growth of around 1.5–1.7% per year, not a plateau in aggregate headcount. (ilo.org) At the same time, some large firms (e.g., HP and other tech companies) are cutting thousands of jobs partly in connection with AI‑driven restructuring, showing that at least some organizations are not keeping headcount flat. (sfchronicle.com) But these aggregate and anecdotal data don’t tell us whether “most” individual organizations have been roughly flat versus modestly expanding.
• Per‑employee productivity from AI: There is credible evidence of productivity gains from AI tools in specific domains—e.g., a 2024 study on GitHub Copilot finds around a 5–6% increase in individual and project‑level productivity for open‑source software development. (arxiv.org) However, it’s not yet clear to what extent such gains have scaled across most organizations or translated into broad, “notably higher” organizational efficiency at the economy‑wide level.

Because (1) the stated time horizon has not yet elapsed, and (2) current evidence does not cleanly confirm or falsify the precise combination of flat headcount for most organizations plus significantly higher per‑employee productivity from AI, the prediction cannot yet be definitively judged. Hence the result is inconclusive (too early).

By November 30 2025, only about 1.5 years have passed since the June 2024 prediction, which explicitly referred to a timeframe of a couple of years (roughly mid‑2026). That window has not yet elapsed, so the outcome cannot be definitively judged. Current data suggests the market has not yet shifted in the way described: surveys show around 80% of organizations pay for subscriptions to tools like ChatGPT or Microsoft Copilot and 63% are using cloud AI APIs, while only 39% are using open‑source models on their own infrastructure and 27% are training proprietary in‑house LLMs. (siliconangle.com) Another survey cited by the Financial Times reports that only about one in eight commercial AI workloads run on open models, with most customers preferring paid, state‑of‑the‑art proprietary systems and even open‑model users typically mixing them with commercial ones. (ft.com) A separate study finds open models account for only about 20 percent of usage and 4 percent of revenue in the AI market despite being up to 84 percent cheaper to operate. (itpro.com) Enterprise spending on generative AI is still increasing overall, with a majority of organizations planning to boost LLM spending, and OpenAI and Google models remaining widely used. (electronicspecifier.com) Netskope telemetry likewise shows that enterprise AI use is dominated by SaaS gen‑AI apps such as ChatGPT, Gemini, and Copilot, along with managed cloud AI platforms, rather than internally hosted open‑source stacks replacing external services. (reddit.com) These indicators imply the prediction has not clearly come true yet, but because the stated timeframe runs to around mid‑2026, it is still too early to call it definitively right or wrong.