Mike Taylor · hedge-fund PM (ex-Millennium under Izzy Englander; healthcare/pharma background, now trading AI and macro long/short) · returning guest of Keith McCullough on Hedgeye's Real Conversations.
“Short euro long gold… is going to be a gigantic trade” next year: gold likes money printing, the only lever left to over-levered sovereigns. Timing waits for bond yields to stop rising (“when I see it, I’ll know it”).
Long select software: as AI compute turns into a surplus, ServiceNow can become the “middleman,” aggregating compute for thousands of embedded clients at a better price. That is free business that deepens integration while the capacity owners lose pricing.
Named with ServiceNow as a compute aggregator: route thousands of clients’ AI workloads through its platform, negotiate the price down and get more embedded. The software side wins as compute pricing falls.
With OpenAI ~70% of the AI market but ~$100B+ of revenue against ~$1T/yr of compute carrying cost; still, “Anthropic is going to get their deal done,” with Musk’s blessing. A shake-out survivor, not a winner of the math.
Pharma internalizing AI to protect proprietary data (“have their own AI in-house”), one more drag on third-party compute pricing. An example, no view on the stock.
Flagged for “a lot of OpenAI exposure” (the lab he expects to be squeezed out) as the host asks what happens if Google cuts capex. A risk flag, not a stock call.
Part of the ~$500B, six-group AI financing; the hosts mock Solomon’s “I believe deeply in these numbers” as narrative in place of math. A financing reference.
His deficit lock-in example: “Coca-Cola is building capacity to service a growing deficit,” so cutting spending would leave excess capacity and imploding pricing. Illustrative, not a stock call.
Named with Bristol Myers as pharma bringing AI in-house rather than sharing data with data centers. An example for the compute-pricing thesis, no view on the stock.
The hyperscaler that showed “an appetite to cut” AI capex, and the market rewarded it. Corporates will fund AI at 3–4% of free cash flow, not the 10% the build-out needs.
3x-levered semis ETF “traded down 17%” in a day: retail and levered-ETF flow whipsawing a semis-minus-software spread that moves 10% every 48 hours. A market-structure reference.
“SpaceX went at $2 trillion”: private holders are 10x+ on self-set marks and want liquidity. Evidence that private money for AI-era valuations is tapped out.
Unsecured bonds at ~12% are “meaningfully distressed” while street models triple its debt in two years. “You are not going to raise that debt because the cost of capital is too high,” and that 12% cost of capital sits inside AI compute’s ~$1T/yr carrying cost.
Protest parties polling “giant” into UK/French general elections from next year will blame the euro and go “back to the franc” (France is ~59% government, overspending 5%). Short euro / long gold, while the US “outruns the slowest bear.”
Supplier to a ~$3T build-out that needs ~$1T/yr to carry against ~$40B of lab revenue left to pay for compute. Chip prices ~6x and memory ~7x inflate the cost, and its financing guarantees are opaque. Negative by thesis, not an explicit short.
Needs ~$600B over two years after a ~$960B round, and “the private money is out.” He reads the regulation push as Altman’s panic button. Too many share the pie, so “they got to get rid of one… it’s OpenAI,” possibly through a bankruptcy that wipes out investors.
The trigger for his bubble call: the OpenAI backlog exploded, “and the bonds said, no,” from March–April. Equity up and credit down is “exactly what happened in 06.”
In one line: A hedge-fund PM (ex-Millennium, pharma background) who calls the AI build-out a bubble on a top-down carrying-cost model: ~$3T of capex needs ~$1T a year, and AI-lab revenue leaves ~$40B to pay for it. Credit flagged it first (Oracle's bonds, CoreWeave unsecured at ~12%). He expects Anthropic to raise and OpenAI to be squeezed out as compute pricing falls, and favours select software as compute middlemen (ServiceNow, Palantir). On macro: sovereign long ends are breaking (short Swiss stocks), Europe is the slowest bear (short euro / long gold next year), deficits are locked into corporate capacity, and inflation falls materially next year as pricing power turns negative. Stance: preserve and protect.
The AI math doesn't close. ~$1.5T of US AI capex in the ground plus ~$1.5T committed = ~$3T by end-2027. Seven-year depreciation is ~$400B/yr, and with power, people, cooling and a 12% cost of capital the all-in cost is ~$1T/yr. OpenAI + Anthropic (~70% of demand) reach ~$100B+ of revenue, leaving ~$40B for compute at an 80% margin. Top-50 S&P FCF is ~$5T, and firms will give AI 3–4% of it, not 10% (2026-SEP-15).
Trust the bonds over the stocks. Oracle's OpenAI backlog exploded but "the bonds said, no," as in 2006. CoreWeave's unsecured debt at ~12% is "meaningfully distressed" while models assume it triples its debt.
Pricing is the next shoe. Pharma is internalizing AI, software aggregators (ServiceNow, Palantir) will push compute prices down, and he reads the regulation press push as a bid for oligopoly pricing power. Private money is out; "Anthropic is going to get their deal done," and OpenAI is the one "they got to get rid of," possibly via bankruptcy. SoftBank is "levered gambling."
The hard part is outside AI. The consumer, housing, auto loans and student loans "stink"; oil corrects only after inventories refill; a white-collar (legal) drag is coming. An end-June index-arb shock broke hedge-fund books, and it's now "the gladiator's arena."
Sovereigns, M2 and Europe. Long ends are breaking globally (JGB 10-year at 3%), and he's shorting Swiss stocks against a 0.58% ten-year. M2 must grow 2–3% to avoid a credit reset while Europe's and Asia's working-age population shrinks ~7% in a decade, so printing is the only lever. The US "outruns the slowest bear": protest-party coalitions in UK/France elections from next year push back toward national currencies. Short euro / long gold is "a gigantic trade" once yields stop rising.
Deficits are now corporate capacity; disinflation next year. Companies built for persistent deficits ("Coca-Cola is building capacity to service a growing deficit," ~$7,000 per person), so spending can't stop. Next year brings "negative pricing power" and inflation "materially down," consistent with host McCullough's nowcast halving by Q2 and 2007-style panic cuts. Retail gets hurt, pushing more money into passive funds; "preserve and protect the pile."
Transcripts
One dated page per appearance — each has its stock table, talking points, and the saved transcript. Newest first.