The Morgan Stanley Tech Momentum Factor Index just suffered its worst unwind in nearly three decades — down 40% in 18 days, a decline with no precedent in data going back to 1998. Yet the S&P 500 barely budged. Equal-weight and small-cap indices were flat. Forty-three companies have reported earnings so far this season, and they’re beating estimates by an average of 16%.
This is not a market in distress. It’s a market rotating, and according to Jordi Visser — a former Morgan Stanley trader who now runs a thematic portfolio — it’s also a market that has structurally changed forever. Speaking on his podcast in mid-July 2026, Visser laid out a thesis that reframes the AI correction not as the end of a bull market, but as the first major speed crash of a new regime driven by AI agents, automated leverage, and concentration risk.
“Markets are moving from manias, panics, and crashes to bubbles, parabolas, and speed crashes — faster up, faster down, and structurally different,” Visser said. The mechanism is already in place: “Every day more and more agents are responsible for trading activity because Robinhood, as of the end of May, is open to agents. Strategies that work harder so you don’t have to.”
The Bank of England has warned about AI agent risk causing market meltdowns. For Visser, that warning isn’t hypothetical — it’s descriptive of events already unfolding across multiple asset classes.
The Correction That Wasn’t: What the Data Actually Shows
The selloff that gripped tech momentum in July 2026 looks catastrophic in isolation, but the broader market tells a completely different story. The key numbers:
MetricJuly 2026 PerformanceHistorical ContextMorgan Stanley Tech Momentum (18-day)–40%Worst since 1998S&P 500 Equal-Weight~0%Rotation, not stressSmall Caps~0%No contagionS&P 500–1.5%MinimalNASDAQ 100 (QQQ)–4%Larger but containedS&P 500 Earnings Surprise+16%43 companies reporting
The thematic portfolio Visser manages correlated nearly 1-to-1 with the momentum factor during the run-up, confirming that the selloff was a concentrated unwinding of the most crowded long positions. “The all-you-can-eat AI buffet has left everyone full and in need of a digestion period,” he said.
Behind the rotation sits an economy running hot. Visser noted that the composite Philadelphia and New York Fed indices are equivalent to a PMI reading of 62 — well into expansion territory — and that JPMorgan and Goldman Sachs both posted enormous earnings beats, with Jamie Dimon describing conditions as “getting close to as good as it gets.”
Visser’s recession signal — the S&P 500’s year-over-year return — remains solidly positive. “We’d have to fall significantly and then bounce” for it to turn negative, he said.
Speed Crashes Are the New Volatility Regime
Visser’s structural argument is that speed crashes are not bugs — they’re features of a market increasingly driven by AI agents executing optimized strategies with embedded leverage. The pattern is visible across multiple assets:
AssetDrawdownTimeframeSilver–40%5 daysBitcoin–40%18 days (historical pattern)Tech Momentum Vol (30-day)92nd percentileJuly 2026Gold 60-day VolRecord spike (except one)2026
“This is only going to get worse,” Visser warned. The vol surface decoupling — tech vol spiking while S&P 500 vol stayed low — creates a portfolio management problem that will make it difficult for leveraged funds to re-enter the AI trade. A position that is seven times as volatile requires a much smaller allocation, forcing systematic strategies to underweight exactly the names they’d want to buy.
The correlation between Visser’s thematic portfolio and the momentum factor approached 1 during the Opus 4.5-driven rally, which he dates to November 2025. By June 2026, the discovery phase had ended. “The firework show is over,” he said. “Agentic AI moves from discovery to digestion.”
The Insatiable Demand Nobody Can Dispute
For all the concern about hyperscaler debt and capex sustainability, Visser is unequivocal: the demand story is intact, and the debt fears are overblown. He points to two data points that he believes settle the argument.
First, the debt-to-equity ratios for Microsoft, Amazon, Google, and Meta are “next to nothing.” The debt they’re issuing funds compute construction, and their combined backlogs and contractual obligations already exceed existing capacity.
Second, and more importantly: “The compute demand is insatiable. Don’t listen to anything you hear from people that argue otherwise. It is insatiable. We will not be there anytime in the next four years where we have enough compute for all the intelligence in the billions of agents that are coming.”
DRAM prices across DDR3, DDR4, and DDR5 are all rising. Micron’s recent earnings showed long-term contract signings that Visser calls a game-changer — so much so that he repurchased Micron in mid-July after selling his last shares in early June. “There was plenty of new good news since I sold,” he said, citing those contracts and rising price guidance.
He contrasts Micron’s position with Nvidia’s. After Nvidia’s June 2025 earnings beat, the stock stopped rallying on good news. “A company can report extraordinary growth and still disappoint,” Visser noted. He believes Micron is earlier in that cycle and may follow a pattern similar to Amazon’s 2013 breakout — a head-and-shoulders neckline break followed by an immediate reversal back above it, the move that originally convinced him to go to Silicon Valley.
On the AGI timeline, Visser elevated a piece by Google DeepMind’s Demis Hassabis from mid-2026: “The artificial general intelligence is probably only a few short years away. When we look back on this time, we will realize we were standing in the foothills of the singularity, nothing less than the dawning of a new age for humanity.”
Visser noted that Hassabis’s 2010 prediction of AGI by roughly 2030 is now being shortened, and called this the most honest and credible AGI timeline from someone not running the company — Sundar Pichai is the CEO. The implication: hardware buildout will accelerate, not slow.
Crypto Decouples and Becomes the Next Beta Trade
The most provocative part of Visser’s thesis is that crypto — not AI infrastructure — will provide the market’s beta over the next 12 months. He rotated a significant portion of his AI stock sale proceeds into Bitcoin, Ethereum, and silver in June and July.
His reasoning rests on three pillars. First, crypto cannot be disrupted by AI because it is built on scarcity and has survived as a “religion” with a moat similar to gold’s. Second, consumer agents will need tokens to transact, and the infrastructure for tokenized assets is proliferating rapidly: JPMorgan, BlackRock, and Goldman tokenized stocks and treasuries in July 2026; Stripe acquired Bridge, a stablecoin infrastructure firm, in October 2025 and is now looking at PayPal. Third, regulatory progress is accelerating: Japan reclassified Bitcoin as a financial asset in July 2026; South Korea wrote crypto into state asset rules; and the US Clarity Act has an estimated 40% chance of passage.
“Bitcoin is the collateral of the future,” Visser said.
The performance divergence from equity momentum is already visible:
AssetJuly 2026 PerformanceRelationship to Tech MomentumBitcoinFlat to slightly positive (above 20-day and 50-day MAs)Previously ~0.9 correlated; now decoupledEthereum+17% month-to-date (best month since August 2025)DecoupledVisser’s 40-name crypto indexOutperforming Bitcoin year-to-dateDecoupled
“If you break the 200-day moving average on the ETH/BTC ratio or on Bitcoin itself, I believe most people, if not all, will be heavily involved in crypto over the next 12 months,” Visser predicted. “I don’t think we’re going to necessarily break through before October or November, but if it does happen, be ready.”
The crypto thesis is reinforced by improving on-chain data. According to supplementary reports, Bitcoin exchange netflows have turned negative, with outflows dominating — a sign that selling pressure is fading. Spot Bitcoin ETFs logged two consecutive weeks of inflows in July after months of relentless outflows. And Glassnode analysis shows that long-term holder realized losses, which spiked above $390 million per day at their cycle peak, have begun to ease.
The Fed: Reformist, Not Hawkish
Visser interprets recent Federal Reserve commentary as dovish, driven by structural deflationary forces from AI and a reform agenda rather than inflation-fighting hawkishness. He cited Morgan Stanley’s assessment: “Worsh is all about reform, not hawkishness.”
The Cleveland Fed nowcast model expects headline CPI to decline to 3.3% for July, with core also declining. The swap market did not budge despite oil backup and Middle East tensions — markets see cuts ahead. Visser noted that the Cleveland Fed nowcast was off by 0.2 percentage points for the previous month, forecasting -0.2% when the actual print was -0.4%, suggesting the disinflationary trend may be accelerating faster than models capture.
“Worsh is rhetorically hawkish about the inflation outcome, but intellectually reformist — he understands the economy,” Visser said. “AI will raise prices, but don’t call it inflation.” He believes July 2026 could see a rate cut if data cooperates, especially with midterm elections approaching.
“It would be a huge mistake for a central banker to be making decisions based on one month’s data, the past,” he added. “Worsh understands the forward-looking data.”
What Consumer Agents Mean for the Next Leg
Visser expects consumer-facing AI agents to arrive within six months — roughly early 2027 — and he believes they will massively increase token consumption, creating a second wave of demand for the infrastructure that the market is currently selling. Apple, which he describes as a “consumer agent story,” rallied on bad news in July 2026, a signal he interprets as early positioning for that catalyst.
The open-source versus proprietary model debate will shape how this plays out. Visser argues that in the medium to long term, open-source models will dominate most enterprises over proprietary frontier models, and AI-native companies will ultimately win over legacy public companies. However, he does not expect this shift to happen in the near term. In the meantime, leading US AI model companies like OpenAI and Anthropic are likely to receive government protection due to the strategic race against China, despite their high capital needs.
The Vera Rubin telescope buildout represents another catalyst on his radar. Visser plans to release a paper in late July 2026 covering approximately 25 new names, many Asian, with step-function revenue potential from the project. His 800-volt DC infrastructure thesis also remains in play.
As the digestion period continues — Visser expects it to last several more weeks with bottoms likely to be tested — the pattern to watch for in memory names like Micron is a head-and-shoulders neckline break followed by an immediate reversal back above it. That was Amazon’s 2013 signal, and if it repeats, it would mark the end of the “Nvidia pattern” of selling every beat. In the meantime, crypto has already begun to decouple from tech momentum, and for Visser, that decoupling is the signal that the next rotation is underway. The structural bull market isn’t over — it’s just changing hands.