Jay Singh — This Is When The Market Rally Ends And Stocks To Beat Downturn

Alamos is a mid-sized gold miner whose shares got hammered when an earthquake hit one of its mines. Singh's point is that the market punished the whole company for damage at a small mine while the flagship operation — the one that actually matters — kept running well and is on track to lift production above a million ounces a year.

He bought into that panic and says the timing was "almost perfect." The wider reason he wants gold miners at all: he thinks real interest rates (what you earn on bonds after inflation) have temporarily peaked, and gold does well when they stop rising.

One of four gold miners he holds together as a single bet: if real rates have peaked, gold rises, and miners rise faster than the metal because their costs are largely fixed. He was briefly underwater on this year's gold buys while the ten-year yield kept climbing; now "the gold miners are up healthily and I think they'll continue to run."

The senior producer in his gold sleeve, held through its recent results. Same thesis as the rest of the basket — this is an interest-rate trade expressed through mining shares rather than a call on any one company's operations.

The quality name in the group, held alongside Alamos, Kinross and Barrick. Owning four miners plus physical bullion is deliberate: it spreads out the single-mine operational risk (an earthquake, a permit, a strike) that would otherwise dominate the outcome of a rates bet.

This is a fund that holds actual gold bars in a vault, so the price tracks bullion rather than any company. Singh holds it beside the miners as the low-drama half of the trade: if he is right that real rates have peaked, the metal alone should work even if a mine disappoints.

Redwood makes large "jumbo" mortgages to creditworthy people who don't have a regular paycheck — doctors, dentists, business owners with 750-plus credit scores — then bundles those loans and sells them on. It is a real, well-run business, and Singh has spoken with management directly.

The opportunity was mechanical rather than fundamental. When an index is rebalanced, funds that track it must sell whatever is being removed regardless of price, and rate fear added to the selling. He bought at "a three handle" — roughly $3 — on a 17% dividend. The stock has since gone to about $4.77, still a 15% yield he believes is sustainable. Buying what index funds are forced to dump is one of his standing methods.

Merger arbitrage: when one company agrees to buy another for cash, the target's shares usually trade a little below the agreed price, because the deal might not close. You buy the target and collect that gap when it does close.

Here Paramount Skydance is buying Warner Bros. Discovery for cash, and the gap is about $5 a share — roughly 19%. Singh thinks it clears antitrust and closes next year. Because the payoff depends on regulators rather than the stock market, this kind of position holds up when the index falls — which is exactly why he wants it heading into an uncertain November.

The same merger-arb structure applied to railroads: Union Pacific is combining with Norfolk Southern, and the gap between today's price and the deal terms is about 12.5%. A wider spread than usual signals the market doubts approval — which is the bet he is taking the other side of.

Datadog sells software that monitors whether companies' systems are running properly. Nothing changed about the business — what changed was who was selling it. Hedge funds piled into AI chip stocks and, to protect themselves, sold software stocks short as the offset. Because AI names are far bigger than any one software company, that hedging crushed software prices mechanically.

Singh bought the wreckage. Datadog "has done very well," and cybersecurity — which he considers the least AI-replaceable corner of software — was his largest position within the basket.

Wix lets people build and host websites for a small monthly fee. The bear case is that AI will let anyone generate a site for free. Singh's counter is about price, not capability: the subscription is so cheap that switching isn't worth anyone's time — "it doesn't make sense for people to replace their web subscriptions with AI." He bought it near the lows and it has rallied.

ServiceNow runs the workflow systems large enterprises use to manage IT and internal processes. Once embedded, it is painful to rip out — that's the "moat" Singh is paying for. He bought it as part of the forced-selling software basket, on the view that a company this entrenched shouldn't be repriced just because someone needed a hedge.

Snowflake stores and organizes company data in the cloud — the layer AI systems have to sit on top of, which makes the "AI kills software" story a strange fit here. Bought alongside ServiceNow out of the same short-hedging sell-off. His honest caveat after the squeeze: he no longer has the confidence to own software "in the same size" he did before.

Babcock & Wilcox builds power-generation equipment — a direct beneficiary of data centers needing enormous amounts of electricity. Singh bought the preferred shares (a higher-ranking, dividend-paying class that behaves more like a bond) as well as the common. The preferreds roughly doubled and the stock rose 40%.

He raises it to make a social point rather than a pitch: people who follow markets closely can convert a visible trend into money, while the average household simply absorbs the higher power and water bills the same trend creates.

TLT holds long-dated US government bonds, so it rises when long-term interest rates fall. Buying it is a straightforward bet that yields have peaked — and he bought last week, "for the first time in years," which is his way of signalling how unusual the call is for him.

He isn't forecasting a collapse in rates, just flat-to-slightly-lower, with one named risk that would break it: a major escalation of the war, not a minor one.

Uber has gone from burning billions a year to generating about $10 billion of genuine spare cash on a $150 billion market value — and he expects $13 billion by 2027 and $15 billion by 2028. At 12 times next year's earnings, down from $100 to the 68 he paid, "it's never traded like that."

The reason it's cheap is the fear that self-driving cars (Waymo) make Uber redundant. His answer is that Uber becomes the demand network those fleets plug into — it signs contracts with them rather than competing head-on. He is explicit that this is a longer-term holding he does not expect to rally soon.

A small position bought recently alongside Uber, and grouped with it as something he doesn't expect to move quickly. The common thread is buying quality that has been left behind while money chased AI.

AppLovin runs the advertising engine behind mobile apps and, increasingly, e-commerce. Singh is honest that this is a call he got "a little bit wrong" so far: his average cost is in the low 300s against a price of 312, leaving him roughly $20-25 a share underwater at one point.

He is holding anyway. Bank of America's worry is that revenue growth slows from 30% to the mid-20s — which he points out is still exceptional. He sees about 50% upside and thinks the stock "could be easily back at 500 over the next 2 years."

Shopify provides the software that online stores run on. He owned it from 120, watched it fall to 97, and it now trades at 158 — but the useful part is the sequence: two blowout quarters in a row produced no rally at all before the third one finally got paid.

He uses it as the honest illustration of this market: being right on the business tells you very little about when you get compensated. "In May, we were eating on Shopify, and it rallied."

Vistra generates and sells electricity. It is his cleanest expression of the theme he thinks defines the second half of the year: AI money moving from chip hype to power and actual returns on investment.

The logic is about recurrence. A data center buys chips once every few years, but it buys power every single year it operates — so power demand is the more durable annuity in the AI build-out. He is also looking at co-location companies (which house other people's servers) for the same reason.

Alphabet reported a strong quarter — revenue up 24%, Search up 17% helped by AI Overviews, YouTube ads and subscriptions both growing — and the stock still didn't rally. The reason is $45 billion of capital spending in a single quarter, which pushed the company's quarterly cash flow to negative $5.85 billion.

Singh's takeaway isn't that Alphabet is bad — "I still like Google" — but that it is no longer the kind of business it was. It used to print cash almost effortlessly; now it consumes it, so the old valuation habits no longer apply. Its stake in Anthropic also flatters its reported earnings without any cash changing hands.

Microsoft is the exception that proves the rule. It rose 13% after earnings for one reason: it said capital spending would stay roughly flat instead of rising again. "Outside of Microsoft, all of the hyperscalers are now burning cash flow."

That makes spending discipline, not revenue growth, the variable that now separates the big tech names from one another — and Microsoft is the only one currently on the right side of it.

Still a name he likes, with a cloud business doing well and its own in-house chips easing the GPU shortage. The caution is shared with the rest of the group: its cloud order book has doubled, but a meaningful chunk of those orders come from AI labs whose money was itself raised from investors — so the backlog is less independent than it looks.

The clean winner from a squeezed middle class. With essentials costing 20-25% more than before the pandemic, middle-income shoppers trade down — "obviously Walmart's doing well" — while other brick-and-mortar retailers, which don't win the trade-down, get hurt on both ends by weak demand and tariffs.

Nebius rents out AI computing power — you pay for time on its graphics chips rather than buying them. Singh's entry signal was concrete and trackable: during the AI sell-off, the rental rates for those chips were rising on Bloomberg, meaning demand was strengthening while the stocks fell.

The quarter validated it — revenue up 454%, profitability flipping from a loss to a $236 million profit, $2.2 billion of operating cash flow, $8 billion of cash, and customers pre-paying about $9 billion. It also has far less debt than CoreWeave. He trimmed into the 34% one-day pop, which is the point: this was sized and treated as a trade.

The same business as Nebius — renting AI computing capacity — bought off the same rising-rental-rate signal, and it also reported strong results. The difference he flags is the balance sheet: CoreWeave carries a lot of debt where Nebius doesn't.

He is explicit that this was bought "for a trade, not for long-term," and that he is selling down the names that have already rallied.

SK hynix is one of the three companies in the world that make the memory chips AI servers depend on. Korean retail investors had piled into leveraged funds tracking it; when the government clamped down, forced selling knocked the whole complex down 45% and left memory names trading at roughly four times earnings.

He bought that dislocation — but labelled it plainly as a trade, not a long-term holding, with an exit already scheduled ahead of November. He also watches its 85% gross margin as the comparison point for whether Nvidia's own margins can hold.

A Taiwanese memory-chip maker bought in the same basket as SK hynix and a DRAM fund, for the same reason: the July panic priced these businesses as if the memory shortage had ended. Same caveat, in his words — "just for a trade, not for long-term."

Nvidia's August 26 results are, in his framing, the read-through for the whole AI complex — so he lays out exactly what to check: the data-center segment (85-90% of revenue) and whether its quarter-on-quarter growth is holding, the transition from Blackwell chips to the next generation, forward guidance covering the next two to four quarters, and whether gross margins stay in the 73-75% range against rising packaging and memory costs.

The bigger reservation is financial engineering. Nvidia is raising $500 billion to help finance purchases of its own chips — money that comes back as revenue. That kind of loop "works in the short term," and he expects Nvidia to keep beating for a couple of quarters. "At some point next year, the market will say — how long is this sustainable?"

Meta is his historical template for over-investment: it has poured money into projects that "failed miserably," most famously virtual reality — then cut costs, recovered, and moved on to the next thing. He thinks AI is a far more real opportunity than VR was, so the pattern isn't automatically bearish.

What he does flag is the social cost that feeds back into markets: Meta is laying people off to fund data centers, which is a large part of why white-collar workers feel worse than the earnings data suggests they should.

This fund holds a basket of software stocks, and he pulls it up as the picture of the call he got wrong. The thesis — that software had been crushed by hedge-fund hedging and would recover — was right. The path was brutal: down 35%, a false bounce, down 15% more, a big rally he sold into, then another 25% drawdown before it finally worked.

The lesson he draws is about survival rather than analysis: being early on a good idea is indistinguishable from being wrong unless you can hold the position, which is why he hedged with a short index position and never used borrowed money.

Atlassian makes the project-tracking and collaboration software many engineering teams run on. He bought it in March in the 70s, watched it fall to the high 50s, and was on the verge of selling at a loss — then it jumped 40% on earnings and now trades at 165.

He sold into that strength and took the gain. The reason is discipline rather than a changed view of the company: after the hedge-fund short unwind lifted every software name at once, he simply doesn't want the same size in software any more.

QQQ tracks the Nasdaq-100. Shorting it — betting it falls — was his hedge, not a forecast: it let him keep the individual software positions he believed in through drawdowns that would otherwise have forced him out.

He covered most of that hedge after the weak August 7 jobs report, reading it as a sign the Fed had less reason to hike. He expects to put risk back down ahead of November.

A 24-year-old former OpenAI researcher wrote a widely-read 165-page paper about where AI was heading, raised money on the strength of it, and built a fund that was long AI infrastructure and short software — with roughly four times leverage on $45 billion of exposure. He was up more than 400% before giving it all back and selling his entire public portfolio to Citadel.

Singh's reading is that the analysis was fine and the leverage was fatal: "if he only leveraged 50% he'd still be running billions today." It is also why the software stocks he owned suddenly exploded higher — when a fund that size is forced to buy back its shorts, the prices go straight up regardless of fundamentals.