NextFin

AI Skeptic Rajiv Jain Embraces the Tech Trade He Once Shunned

Summarized by NextFin AI
  • Rajiv Jain of GQG Partners reversed his AI-bubble stance, moving his U.S. fund from under 3% tech exposure to overweight semiconductor stocks ahead of September 2026.
  • The Philadelphia Semiconductor Index fell 21% in July 2026, its worst month since October 2008, erasing about $2.2 trillion in market value before the pivot.
  • GQG reported net outflows of $15.1 billion in the first half of 2026, with funds under management falling to $156.0 billion from a peak of $172.9 billion.
  • Jain's new thesis bets suppliers earn before customers fail, distinguishing realized chipmaker earnings from hyperscaler capital-expenditure promises.

NextFin News - Rajiv Jain, the chairman and chief investment officer of GQG Partners, has bought the artificial-intelligence trade he spent two years calling a bubble worse than the dot-com bust. After cutting technology exposure in his U.S. fund to less than 3% while the category average sat near 30%, after deploying a team of six former investigative journalists to test whether the GPU shortage was real, and after defending the call as billions of dollars left his funds, Jain moved his firm overweight into semiconductor stocks in the weeks leading into September 2026. The reversal does not mean the bubble thesis is dead. It means the most disciplined bear in the room has concluded that the picks-and-shovels suppliers will get paid before the customers who hired them go broke.

The Reversal, and What It Is Not

The pivot is specific, and the specificity is the story. GQG has moved into semiconductors - the chip designers, foundries, and memory makers that supply the AI buildout - rather than into the hyperscalers whose capital expenditure is the source of Jain's original alarm. His argument rests on earnings: the chipmakers are already delivering the cash flows that the infrastructure spenders are only promising.

The timing is the second half of the story. Jain did not chase the top. The Philadelphia Stock Exchange Semiconductor Index plunged 21% in July 2026, its worst month since October 2008, erasing roughly $2.2 trillion of market value; Taiwan Semiconductor Manufacturing fell 15% in the month and Micron Technology dropped 29%, its steepest monthly decline in more than a decade. The index had doubled in the first six months of the year before giving it back. Buying into semiconductors after a one-in-18-years drawdown is a different proposition from buying them at a record high - and it is the difference between a valuation-driven entry and a momentum-driven surrender.

That distinction matters because the headlines have framed the move as capitulation. They are not wrong to notice it. But a capitulation follows price; a repositioning follows a change in the underlying math. Jain's claim is the latter: that the supply side of AI - memory, foundry, and equipment - is trading on realized earnings from a genuine shortage of high-bandwidth memory and advanced packaging capacity, while the demand side - the cloud platforms funding the buildout - is trading on promises that the accounting cannot support.

The Bill for Being Early

The reversal comes after GQG paid heavily for its discipline. The firm's four main strategies underperformed their benchmarks in the first half of 2026, a result it attributed in an exchange statement to "the positioning of our portfolios in the current market environment." The statement also promised continued flexibility: "We review portfolio positioning daily and will adjust exposures as opportunities and risks evolve," it said, adding that the firm's focus "remains on compounding client capital over a full market cycle with disciplined attention to downside risk."

The market charged for that discipline in cash, not just in relative returns. GQG reported first-half fiscal 2026 earnings of 8 cents a share on net revenue of US$397.2 million, with operating margin holding at 76% - evidence of the capital-light economics of asset management. But net outflows of US$15.1 billion over the period were only partly offset by US$7.2 billion of portfolio gains, leaving funds under management at US$156.0 billion as at June 30, 2026. The firm's asset pool has since fallen further: from a peak of US$172.9 billion on the eve of the Middle East conflict in late February, it had declined by US$23.7 billion - nearly 14% - by mid-September, with the renewed regional tensions disrupting the very tech pivot that was supposed to stabilize flows.

For an ASX-listed manager whose valuation is a multiple of flows, the pressure was financial as well as reputational. Eight consecutive months of net outflows, totaling almost US$20 billion by late February, had already tested client patience before the pivot. A manager cannot be a contrarian against his own clients indefinitely; at some point the contrarian position becomes the liability, not the asset.

"One of the lessons of navigating these kinds of bubbles is you're either early or late. If you're late, the losses are horrendous," Jain said.

The line was delivered while he was still short the trade. Read after the pivot, it carries a second meaning: being early cost GQG roughly nine months of underperformance, but being late to a semiconductor cycle that had already rewarded its participants would have cost more.

Why the Bear Thesis Was Not Stupid

It is important, for the sake of honesty, to reconstruct the bear case at full strength - because if it was wrong, it was wrong for interesting reasons, and those reasons have not disappeared.

Jain was never anti-technology. In the first quarter of 2023, GQG was the single largest institutional buyer of Nvidia stock, which then traded at 60 times forward earnings. "Until the summer of 2024, some clients would say, 'Don't you have too much in tech?' We started getting worried," Jain said. "This year, almost all the data began pointing in the other direction. We grew concerned that the data doesn't jibe with what people are saying."

The data he gathered was unusually granular. When Wall Street spoke of a severe GPU shortage, GQG's team of former investigative journalists tried to order a parcel of up to 5,000 graphics processors. They found numerous resellers offering to ship the chips within a week at a massive discount. On Nvidia's own website, the H200 was listed at more than $40,000; one reseller quoted $25,900 - a discount Jain put at 50% to 60%. If a shortage were genuine, Jain argued, that spread would not exist.

The valuation comparison he drew to the dot-com era was equally concrete. In early 1999, the S&P 500 traded at a low-20s earnings multiple, with five-year EPS growth above 20%. For the five years ending 2026, S&P 500 EPS growth was expected to be around 8%. Then, Microsoft was growing earnings above 40% and revenue above 30%. "I don't know of any company that is growing EPS near these rates, except Nvidia," Jain said, "and we know that is cyclical." More than half the S&P 500, he noted, is now directly or indirectly linked to the AI trade: "The S&P 500 itself is a growth stock."

The mechanism he identified was accounting, not mood. The hyperscalers - Microsoft, Alphabet, Amazon, Meta Platforms, Oracle - are funding the buildout with free cash flow that is increasingly strained by data-center and GPU spending. Their reported earnings quality, Jain argues, is being supported by extending the useful lives of servers and chips to five or six years, versus the two to three he believes is realistic, and by adding large stock-based compensation back to cash-flow metrics as if it were a non-cash item. Those are falsifiable claims about depreciation schedules and cash-flow definitions. They do not become false because a chipmaker prints a good quarter.

The Contradiction Inside the Pivot

Here is the tension the pivot creates, and it is the heart of the matter. Jain's original thesis was that AI capital expenditure would fail to earn an adequate return, and that when the shortfall became visible, the entire complex - including the industrial and materials suppliers to data centers - would reprice. His new position is that the semiconductor suppliers will keep earning through the transition.

Both positions can be true only if timing breaks the link between them. The semiconductor cycle runs on its own clock - memory pricing, foundry utilization, equipment order books - which is only loosely coupled to the hyperscaler capital-expenditure cycle. High-bandwidth memory prices surged in 2025 and were expected to rise further into 2026 on genuine AI-driven demand, independent of whether Microsoft ultimately earns a return on its GPU fleet. A fund can be long the cycle and short the bubble at the same time, for a while.

But the coupling does exist, and it is the risk. If the customers who fund the semiconductor up-cycle eventually cut spending because their projects do not pay, the cycle turns - and it turns hard, because memory and foundry capacity is expensive to idle and expensive to restart. Jain's bet, stripped of its elegance, is that the capex downturn will arrive after the suppliers have had time to earn, and that the earnings will be visible enough to carry the stocks through the transition. That is a timing bet dressed as a valuation bet. And timing bets are where value investors go to die.

There is also the question of whether the pivot is a change of mind or a change of circumstance. The July semiconductor drawdown improved the risk-reward of the very stocks Jain had avoided. A 21% one-month decline does not just make a sector cheaper; it makes a U-turn defensible. Investors are entitled to wonder how much of the reversal is conviction and how much is the market having already done part of the contrarian's work for him.

The Strongest Case Against Him

The counter-thesis is not that Jain's math is wrong. It is that his behavior is indistinguishable from the crowd he criticized. For two years he warned that the AI trade was driven by narrative rather than fundamentals, that Wall Street was an echo chamber where the sellside talks to management, management talks to the buyside, and the buyside talks to the sellside, "and it's happily ever after." Now he is buying the sector that every momentum manager has owned, after that sector's worst month since the global financial crisis - the exact moment when the narrative shifts from "AI is a bubble" to "AI was a buying opportunity."

The evidence for this reading is not trivial. The semiconductor rally of 2025 and early 2026 was led by memory - Micron, SK Hynix, Samsung Electronics - on a genuine pricing cycle in DRAM and high-bandwidth memory. But by mid-2026 the rally had cooled: a Taiwan Semiconductor outlook in July helped trigger a broad chip selloff, and analysts noted that the momentum was fading even as the companies reported record profits. Buying into a sector where the earnings are peaking and the multiple is contracting is how the discipline that made the reputation breaks the reputation.

Jain's answer would be that earnings, not momentum, justify the exposure, and that the valuation gap between semiconductors and the rest of the market has narrowed even as the AI narrative has widened. The falsifying test is specific and observable: if semiconductor earnings growth decelerates for two consecutive quarters while hyperscaler capital-expenditure guidance holds or rises, the "the suppliers are safe even if the spenders are wrong" thesis breaks - because it would mean the suppliers are being paid by spending that is not earning its keep, a chain that cannot hold. The confirming signal runs the other way: if hyperscaler free cash flow compresses materially while capex guidance is cut, the original bubble thesis is vindicated even if the semiconductor trade profits along the way.

What to Watch

Three signals will determine whether this pivot is remembered as foresight or as capitulation.

First, memory and foundry pricing through the next two quarters. Sustained price strength in high-bandwidth memory and advanced packaging confirms a genuine supply-demand cycle; a roll-over confirms the demand was pulled forward by a buildout that is slowing.

Second, hyperscaler capital-expenditure guidance versus free-cash-flow realization. The gap between the two is the bubble's pressure gauge, and it is measurable in quarterly filings. When the guidance stays elevated while the cash flow deteriorates, the accounting adjustments Jain warned about become visible to everyone, not just to him.

Third, GQG's own flows. A stabilization would suggest clients are willing to follow him into the reversal; continued outflows would suggest the credibility damage from being early - and the suspicion that the pivot is performance-chasing - has not been repaired. The firm's own shares, trading near the lower end of their 52-week range after the August drawdown, are a daily referendum on that question.

The base case is that Jain is half right: semiconductors grind higher on real earnings while the broader AI complex grinds lower on unrealized promises. The upside case is that the capital-expenditure cycle proves more durable than his models allow, and the entire complex rerates. The downside case is the one he warned about for two years - a capex strike that takes the chipmakers down with the spenders, proving that you cannot be long the picks-and-shovels and short the mine forever.

Rajiv Jain spent two years telling investors that the AI trade was a bubble waiting to burst. He may yet be right. But the market's most expensive lesson is that being right too early looks exactly like being wrong - and the bill arrives in flows long before it arrives in prices. The difference this time is that he has decided to collect the payment on the way down.

Explore more exclusive insights at nextfin.ai.

Insights

What was Jain's AI bubble thesis?

How did GQG test the GPU shortage?

Why compare AI to the dot-com bust?

What accounting tricks worried Jain?

Who is Rajiv Jain at GQG Partners?

How much tech exposure did GQG hold?

Did semiconductor stocks crash in July?

How much capital left GQG funds?

Why buy chips but not cloud firms?

When did GQG switch to chip stocks?

What drove the July 2026 chip selloff?

How did GQG earnings look in H1 2026?

What signals will prove Jain right?

Can suppliers survive a capex cut?

What downside risks face chip stocks?

Will GQG client flows stabilize soon?

Is Jain's pivot conviction real?

Why is timing risky for value funds?

Did Jain join the crowd he mocked?

What breaks the supplier safety thesis?

Search, ask or analyze anything...