Summary

Cathie Wood’s central macro call is that 7%-plus real global GDP growth is conservative, not a bull-case flourish. Global growth stepped from roughly 0.6% in 1500–1900 to 3% across the railroad, electricity, telephone and internal-combustion revolution; she expects robotics, energy storage, AI, blockchain and multi-omic sequencing—15 technologies converging across five platforms—to drive another step change. “It’s nothing that anyone living today has seen before.”
Falling technology prices need not shrink the economy because lower costs can trigger explosive unit demand and convert unpaid activity into measured commerce. Wood expects inflation below 2% and “heading negative” within a year if productivity rises and unit labor costs decelerate; Truflation was already at 1.2% by her cited measure. Robots also move cooking, cleaning and driving children from unpriced household labor into GDP: “We’re going to unlock a lot of that.”
Collapsing inference costs may not weaken AI infrastructure demand because the appetite for intelligence is effectively unbounded. Diamandis described cognition as being commoditized at roughly 99% per year, while Wissner-Gross argued that users will spend savings on longer reasoning loops or parallel agents—100 attempts can improve the odds that one succeeds beyond an agent with an 80% success rate. OpenAI’s challenge is monetization: Wood cited roughly 900 million users and a prospective advertising price near $60 per thousand views or engagements versus Facebook’s $20, while Gemini can use Google’s cash flows to compete without matching that strategy.
China’s open-source AI mobilization is the strategic counterweight to the US lead in applications. Wood said China has moved ahead of the US in open source after DeepSeek’s success and Llama 4’s weak showing, while investment remains around 40% of Chinese GDP versus just over 20% in the US. Ismail argued the US can still win at the application layer; Wissner-Gross countered that concentrating core research inside a few closed labs throttles the number of ideas, even if China’s open models flow back to US developers.
Wood retained ARK’s $1.5 million Bitcoin bull case for 2030, but its composition has changed. Stablecoins—especially Tether in emerging markets—captured part of Bitcoin’s anti-confiscation use case, subtracting an estimated $200,000–$300,000, while gold’s doubling and an intergenerational shift toward “digital gold” restored support. She sees Bitcoin hedging both inflation through its 21 million cap and catastrophic deflation through self-custody without counterparty risk: “Its cause is freedom, financial freedom.”
ARK expects disruptive-innovation equities to compound at 35% annually over five years and sees convergence creating a possible $100 trillion company by 2030. Wood named ARKK as the flagship expression and Tesla as the leading corporate candidate because its road, energy, robotics, manufacturing, X, Neuralink, SpaceX and Boring data could reinforce one another. Her attack on conventional indexing is blunt: benchmark leaders encode past success just as disruption is preparing to reorder the market.
Power is the binding infrastructure trade, with cumulative global investment needing to reach $10 trillion by 2030. China was building about 28 large nuclear reactors while the US was building no large reactor; Wood argued that uninterrupted nuclear learning curves could have left US electricity 40% cheaper. She also highlighted first-year depreciation for qualifying US manufacturing structures whose construction starts before the end of 2028 as fuel for reindustrialization and an approaching “economic boom.”
Robotaxis could destroy today’s auto-volume model even while creating enormous platform cash flow. ARK calculates that 24 million highly utilized vehicles could cover all US urban miles versus roughly 400 million owned cars today; Tesla could eventually price rides at $0.20 per mile, against Uber’s rise from $2.00 to $2.80, with costs around 50% below Waymo’s. The investment distinction is between manufacturers built around AI, batteries and automated factories and legacy assemblers whose “DNA is not right.”

Deep dive

1. Five converging platforms make 7% growth the conservative case

Wood opened with an admission unusual for a forecaster already known for aggressive projections: “AI is moving faster than we expected.” ARK’s five-year Big Ideas work began in 2017, inspired by Mary Meeker’s data-heavy internet reports but extended from documenting recent history into explicit investment-horizon forecasts.
Wright’s law supplies the operating framework: every cumulative doubling of units produced should drive a technology-specific, consistent percentage cost decline. That shifts the research question from time, as in Moore’s law, to what might impede unit growth.
The historical comparison carries Wood’s 7% call. ARK estimates real global GDP growth at roughly 0.6% from 1500 to 1900, then approximately 3% for the next 125 years as railroads, telephony, electricity and internal combustion diffused through the economy.
Now robotics, energy storage, AI, blockchain and multi-omic sequencing encompass 15 technologies that increasingly reinforce one another. Blundin presented the skeptic’s “3% rut”; Wood replied that living memory contains no comparable multi-platform revolution and that conventional sector-siloed research cannot see technologies “permeating every one of them and blurring the lines.”

2. Orbital data centers expose how quickly convergence rewrites models

ARK’s open-source SpaceX model, built with Mach33, initially omitted orbital data centers because the use case barely figured in the discussion when the model was released, probably around the middle of the prior year. The team was already returning “to the drawing board”—a concrete example of forecasts being overtaken by convergence.
Reusable rockets should move down a Wright’s-law cost curve; Salim said the decline was well into the 20% range per cumulative doubling, though he was not certain of the figure. He contrasted that with industrial robots, where costs decline approximately 50% per cumulative doubling. A new orbital-compute use case raises launch volumes, which accelerates learning and lowers costs for the next use case.
Blundin said he entered a conversation with Elon Musk only “half believing” in orbital data centers and left sold. His stack-level argument included roughly 50% margin at TSMC and 80% at NVIDIA, possible vertically integrated fabrication, cheap raw materials and solar panels said to be six times more efficient in space.
Wood said ARK typically assumes vertical integration, especially with Musk’s companies, and agreed the model could be conservative on some costs beyond launch. Wissner-Gross pushed the extrapolation toward Dyson swarms and lunar disassembly; Wood gave an appropriately uncertain answer, saying ARK had modeled Tesla, Optimus and Boring reaching Mars but that its space analysts needed to study the 50-year question. Diamandis identified orbital debris as the nearer-term showstopper.

3. Good deflation can coexist with explosive real growth

Blundin’s challenge was direct: a NASA shuttle launch cost about $600 million, SpaceX brought that near $60 million and another 10-fold decline was conceivable—so how can collapsing prices increase GDP? Wood’s answer was Jevons-like: “The other side of costs coming down is, of course, explosive unit growth.”
Wood cited Truflation’s real-time basket of 10,000 items at 1.2%, versus the Federal Reserve’s focus on readings around 2.5%–3%. Conditional on accelerating productivity and decelerating unit labor costs, alongside falling gasoline and rents, she expects inflation below 2% within a year and then “heading negative.”
Blundin argued that GDP can misread genuine welfare: curing breast cancer could eliminate costly radiation and chemotherapy, mechanically reducing measured spending while creating immense value. Wood supplied the opposite measurement effect—purchased robots will monetize driving, cooking and cleaning that households previously performed for no recorded income.
Ismail’s Uber example made the demand response tangible. Venture investors modeled the company against San Francisco’s roughly $500 million taxi market and missed that ridesharing would quadruple the category while taking 80% of taxi share. Lower friction created trips that did not previously exist rather than merely reallocating existing fares.

4. Near-zero inference costs still leave near-infinite demand

Diamandis described cognition as being commoditized at roughly 99% per year, while the chart showed collapsing inference costs. Wissner-Gross’s response was: “The demand for intelligence is essentially infinite,” because cheaper inference invites longer reasoning loops, more agents and brute-force parallelism.
An agent completing long-duration tasks with 80% success may look unemployable in isolation, Blundin joked. But launching 100 agents raises the chance that one finds a solution; declining cost therefore converts a quality limitation into additional compute demand rather than simply reducing the bill.
Wood said OpenAI was moving toward advertising, commerce and robots, with a prospective advertising price she understood as about $60 per thousand views or engagements, versus roughly $20 at Facebook. Its 900 million users provide scarcity initially, but ARK’s consumer analysts worry Gemini can take share without copying the model because Google can subsidize it.
Ismail recalled a possibly off-camera mandate to find $75 billion of advertising revenue within two years—perhaps 18 months—from zero. Wood compared that with Amazon reaching about $50 billion after roughly seven years and suggested OpenAI may eventually need to “focus, focus, focus” rather than pursue many deep objectives simultaneously.

5. Personal agents turn AI capability into culture—and operational risk

The panel’s live example was the open-source personal agent first called ClawdBot and then corrected to Moltbot after trademark issues with Anthropic, complete with a lobster mascot. It connected with email, social accounts and laptop data, performed work overnight and returned in the morning “like an eager employee or intern.”
Wood said ARK’s lead AI analyst became visibly better organized after one weekend with it. Wissner-Gross compared the phenomenon with Suno, Sora and Arduino: the important effect is cultural as well as technical, because people can show friends what they built overnight and pull new users into experimentation.
Diamandis supplied the key warning: the same access that makes a personal agent powerful could “scramble your entire computer in two seconds.” Its rapid spread illustrated individual agency and the risks of unconstrained access.

6. AI should punish benchmark hugging before it perfects markets

Wood tied real wealth growth to technologically enabled productivity rather than asset-price inflation. The PC, Microsoft and internet produced only a preview: productivity recovered from near-zero or negative readings in the 1970s and early 1980s, financial markets boomed, and inflation fell because growth increased supply rather than merely demand.
Asked for the better progress measure, she chose gross national income over GDP or per-capita productivity. If productivity is underestimated—she cited the measured rate at about 2%—then real GDP is understated and inflation overstated; widening output-versus-income discrepancies leave policymakers vulnerable to mistakes.
ARK projects disruptive-innovation equities compounding at 35% annually for five years, with ARKK spanning all five platforms. Diamandis cited approximately 31%–33% annualized performance over the preceding two years; Wood said the three-year figures were moving toward the target, though returns would need to exceed 35% later to achieve the full-period average.
Her view is that “the market’s never been more inefficient,” as post-2000 and post-2008 risk aversion pushed capital toward S&P 500 and Nasdaq incumbents selected for past success. Prediction markets and AI-assisted original research could revive genuinely active investing; the first-order effect, she argued, is to “destroy anyone that looks like a benchmark.”

7. China has moved ahead in open source while the US retains application leverage

Wood traced China’s open-source strength partly to Western software companies withdrawing over intellectual-property theft. DeepSeek showed how fully China had exploited the opening, while Llama 4 “falling flat” led Wood to say China was ahead of the US in open source.
Chinese investment, including property, remains around 40% of GDP versus a little over 20% in the US. With property deflating, Wood interpreted the sustained ratio as evidence that Xi Jinping’s emphasis had shifted from “common prosperity” toward “new productive forces”—technology receiving enormous capital allocations.
Healthcare was the concrete proof point. Diamandis cited Insilico Medicine’s Hong Kong listing as 1,200 times oversubscribed, while Ismail noted that China was running more clinical trials than the West. Wood attributed part of that lead to stricter US regulation; Diamandis said the new FDA commissioner was lowering barriers, and Wood said regulation was changing.
Ismail argued the contest would be won at the application layer, where Silicon Valley dominates apart from examples such as TikTok and perhaps Spotify; he conceded energy was a major vulnerability. Wissner-Gross replied that moving core algorithms into a few closed labs throttles idea flow, although dangerous Chinese open-source advances also return immediately to US developers.

8. Bitcoin’s $1.5 million case survived, but stablecoins changed the math

ARK’s 2030 Bitcoin bull case remains $1.5 million. The major negative revision inside the model is that stablecoins—especially Tether in emerging markets—now provide dollar-backed protection against inflation, devaluation and seizure that ARK had expected Bitcoin to supply, subtracting roughly $200,000–$300,000 from the target.
The offset is gold: it had doubled over two years and significantly outperformed Bitcoin over the previous year. Although Bitcoin and gold’s 2020–2025 correlation was only 0.14, Wood said gold led Bitcoin in the last two cycles and expects younger recipients of intergenerational wealth transfers to favor “digital gold.”
She attributed recent weakness to the October 10 “flash crash,” described as a Binance software glitch that triggered automatic deleveraging among highly leveraged participants. About $28 billion was in positions caught offsides, and ARK was hearing that the liquidation overhang had been largely cleared—hence expectations for another “big, big run.”
Diamandis pushed back that Bitcoin had not visibly behaved as the promised inflation hedge. Wood cited an approximately 360% gain from the late-2022 bear-market bottom, its 21 million limit and supply growth falling from 0.8% toward 0.4%; under deflation, self-custody eliminates counterparty exposure, making Bitcoin a hedge against financial-system failure as well.

9. Digital assets could open private markets without immediately replacing trust

ARK sees digital assets reaching $20 trillion in market value—roughly the current US economy and comparable to the US equity market in 2010. Smart contracts and enabling legislation broaden the thesis beyond Bitcoin as currency or store of wealth.
Asked whether ICOs could replace IPOs, Wood pointed to Robinhood’s crypto-savvy infrastructure and efforts to distribute ownership of large private companies, alongside ARK’s interval fund. Her hedged forecast was that an intermediary-led version of the originally decentralized vision is “very possible” within three years.
Blundin’s counterweight was institutional trust. Private securities can trade indefinitely during calm periods, but crises send capital back toward SEC oversight, GAAP accounting and US public-market liquidity; even a16z at roughly $90 billion and General Catalyst at $60 billion remain small beside the trillion-dollar pools an AI build-out may require.

10. Convergent proprietary data could produce a $100 trillion company

Wood said a $100 trillion company could exist by 2030 and named Tesla the leading candidate, potentially through combinations. Her mechanism was proprietary data convergence: Tesla owns “the language of the roads,” while Neuralink, SpaceX, X and Boring contribute multi-omic, orbital, social and underground-infrastructure data unavailable elsewhere.
She had not expected SpaceX to go public because it did not appear to need to do so and Musk’s public-market experience with Tesla was hardly welcoming. If it does list, she suggested the reason would be the orbital-data-center opportunity.
Blundin recalled Musk rejecting the invitation to claim that his companies were deliberately designed to converge: “It’s totally luck.” The episode’s resolution was that AI is forcing formerly separate assets together; ARK’s Tesla work benefited because robotics, energy-storage and AI analysts collaborated instead of leaving the company to an auto analyst versed in human-driven combustion vehicles.

11. Nuclear’s interrupted learning curve left power structurally expensive

Wood’s starting proposition was that “economic activity is energy transformed.” Major economies are using energy more efficiently, but progress still requires more energy; treating energy itself as bad amounts to asking society to return to “the dark ages.”
China looked roughly half as energy-efficient as peer economies on the chart, although Wood cautioned that this was an exaggeration. It was building about 28 large nuclear reactors simultaneously. The US was building no large reactor, although it still had more nuclear plants and nuclear generated approximately 20% of US electricity.
Regulation in the US and Japan during the 1970s reversed nuclear construction-cost declines that had tracked Wright’s law. Wood estimated that if the learning curve had continued, US electricity would now cost 40% less; she therefore wants large, medium and small nuclear systems, all represented in ARK’s venture investments.
Wissner-Gross asked whether nuclear regulation explained the economic rupture around 1971. Wood assigned the larger role to closing the gold window, monetary policy governed by “human frailty,” wage-price controls and broader regulation; oil prices then quadrupled, while nuclear’s shutdown epitomized the same loss of discipline.

12. A $10 trillion power build-out creates opportunities below the headline names

ARK estimates cumulative investment in global power must rise to $10 trillion by 2030. Wood’s summary was categorical: “There are gonna be trillions of dollars invested into AI everything,” spanning generation, data centers, chips, storage, interconnects and supporting infrastructure.
She highlighted US tax treatment as a reindustrialization catalyst: qualifying manufacturing structures whose construction starts before the end of 2028 can be depreciated fully in their first year of service rather than over 30–40 years. The resulting refunds can fund R&D or lower prices, supporting her forecast of an “economic boom in the next few years.”
Boom Supersonic supplied the best venture example. Its difficult aircraft-and-FAA path became a generator business with years of backlog, turning reusable engine capabilities toward AI-related power demand; for Blundin, that showed both how hidden stack components can produce 10- or 100-fold gains and why a great team matters when convergence enables rapid pivots.

13. Robotaxi utilization breaks the economics of car ownership

Uber now accounts for about 1% of US urban miles, Wood said, yet ARK calculates that only 140,000 highly utilized vehicles could serve that volume. Covering all US urban miles would require approximately 24 million cars, against roughly 400 million vehicles owned and 15 million new autos sold domestically each year.
ARK expects Tesla to become the largest robotaxi platform and Waymo the second. Waymo had fewer than 3,000 US vehicles and depended on Zeekr, Hyundai and other suppliers; Tesla’s vertically integrated system should eventually carry a cost structure around 50% lower.
Uber’s average price rose about 40% in four years, from $2.00 to $2.80 per mile, partly through surge pricing. ARK corroborates Tesla’s claimed ability to price at $0.20 per mile at scale; the enormous interim “price umbrella” could let Tesla charge less than incumbents while producing exploding cash flow.
Ismail added an asset-light route: owners could place personal Teslas into the network, echoing Uber’s rapid scaling. Those vehicles would also be mobile inference engines and energy-storage devices that charge or discharge around grid utilization—robotaxis, AI compute and distributed power becoming one system.

14. Automated factories divide the future auto sector from legacy assemblers

Blundin said visiting a Gigafactory changed his view of Musk’s aversion to suppliers. Vertical integration is not merely control: exponential demand cannot be met if one externally sourced component constrains the chain, whereas raw aluminum, chips and reconfigurable robots can feed an internally planned production system.
Wood described Musk’s realization that he was “a manufacturer of factories” as an important ARK insight. Legacy automakers order seats, chassis and drivetrains from third parties; Tesla’s general-purpose factories can redirect capabilities toward vehicles, satellite manufacturing and other robotic products.
Her verdict on incumbents was severe: after pulling back from electric vehicles, they are trying to enter robotaxis despite DNA rooted in combustion engines and human driving. A cumulative production doubling for mature combustion technology might take, she estimated, around 100 years, while EVs continue descending battery and manufacturing learning curves.
The panel preserved a useful disagreement: Ismail could not see today’s automotive industry surviving, while Blundin expected the sector to expand by evolving into robots of many shapes. Blundin reconciled them—the sector may become larger than ever, but companies bound to unions, pensions, suppliers and legacy jurisdictions may be unable to make the pivot.

15. Autonomy reaches delivery as the debate moves beyond labor

The slide put fully autonomous delivery at roughly four million deliveries per year. Zipline’s defining example began with medical deliveries in Rwanda; Wood and the panel credited it with reducing pregnancy-related maternal bleed-out mortality by more than 50%.
Diamandis cited Wing and Singularity University spinout Matternet, and also named Starlink, Meituan and Coco robots in discussing ground delivery. Three-dimensional airspace offers more capacity than streets, but Blundin identified noise as the likely constraint: a materially quieter drone could be a category-defining advance.
Wissner-Gross closed by asking whether automation could eventually substitute not only for labor but for capital itself. Wood said blockchain would transform financial infrastructure but hesitated to call capital “immortal”; Ismail proposed the higher-order progression as money to information to intelligence, while Diamandis added directed intelligence or purpose. Everyone conceded that measurement, exchange and monetization remain unresolved.

Market Overview — February 3, 2026 Ben Horowitz and David Solomon: The Sweetest Macro Spot in 40 Years