NextFin

UniSuper's Pearce Rejects AI Bubble Fears but Draws a Hard Line at 'Priced to Perfection' Deals

Summarized by NextFin AI
  • UniSuper CIO John Pearce is not an AI skeptic but a disciplinarian: he keeps the fund structurally overweight U.S. tech yet refuses to buy Firmus because its $44 billion valuation is priced to perfection.
  • The Nvidia- and Blackstone-backed Firmus withdrew its ASX IPO on October 9, 2026, after market conditions rejected a $30.6 billion equity / $60 billion enterprise value deal that tripled its paper price tag in two months.
  • Pearce frames the current pullback as a valuation correction, not a price correction: the S&P 500 trades just below record highs while multiples stop expanding as the 10-year U.S. Treasury yield sits at 5.15 percent.
  • The falsifying signal is quantifiable: if the S&P 500 falls more than 10 percent from its October 2026 peak while yields stay above 5 percent, the no-bubble thesis fails and the correction becomes earnings-driven.

NextFin News - John Pearce, chief investment officer of UniSuper, is not an artificial-intelligence skeptic. He runs Australia's fourth-largest superannuation fund, keeps it structurally overweight U.S. technology, calls the AI capital-expenditure super cycle a genuine force in the economy, and says the broad share market is not in a bubble. But he will not buy Firmus. The Australian data-centre company's planned float, which Pearce said would raise about $7 billion and value the business at around $44 billion, fails a simple test.

"We think that Firmus indeed has a compelling story. It just doesn't have a compelling valuation. At the current valuation, it's priced to perfection."

The distinction between an AI skeptic and an AI disciplinarian matters because it is being missed. Pearce's October 2026 investment update, recorded on October 7, used the Firmus initial public offering to draw a line that separates enthusiasm for the technology from discipline on price. UniSuper will still end up holding Firmus shares through its index strategies, and external managers may buy them, but the fund will not be participating in the IPO in any meaningful sense. The objection is not to the business. It is to the price, and to the financing structure that the price implies: Pearce said he is "quite concerned that Firmus will have to keep coming back to the market for a lot more debt and more equity to fund its expansion plans."

The Situation: A Disciplinarian, Not a Doubter

The timing was pointed. The update's own editor's note acknowledged that between the recording and its release, reports had emerged that the Firmus IPO "has not been well received" and "may be shelved" unless it was materially repriced. It was. The Nvidia- and Blackstone-backed company withdrew its ASX listing application on October 9, 2026, after what co-founders Oliver Curtis and Tim Rosenfield described as market conditions that did not reflect the business. The deal would have raised about $5 billion at A$11 per share, giving an equity valuation of $30.6 billion and, with roughly $30 billion of debt, an enterprise value of about $60 billion — a deal that deal-flow trackers said would have ranked as the fourth-largest public offering globally that year. A company valued at $10.5 billion in an August fundraising round had, in two months, tripled its price tag on paper. That is the kind of repricing the market refused to underwrite.

Pearce's stance is more nuanced than the headline suggests. He is not calling for a broad retreat from AI. In the same update he rejected the idea that share markets have become unhinged from bonds:

"Has the share market become unhinged, disconnected with the bond market? Are we actually seeing signs of a bubble that's going to burst? I don't believe so."

His read is that the correction so far has been "a correction in valuation rather than price." As of mid-October, the S&P 500 was trading just below its record high while Australia's benchmark sat about 6 percent below its peak. The index had not collapsed; the multiples had simply stopped expanding.

That framing — valuation correction, not price correction — is the crux of the entire argument, and it deserves to be taken literally. In a price correction, earnings estimates fall and the market reprices risk. In a valuation correction, the earnings story is intact but investors are no longer willing to pay the same multiple for a dollar of earnings because the risk-free rate has risen. Pearce is arguing the latter. It is a bullish position on corporate profits and a cautious position on what those profits are worth today.

Why This Is Not 2000 — And Where the Froth Actually Is

The central question Pearce's position forces is whether the AI boom is a structural regime shift or a cyclical mania. His answer, built across three updates in 2026, is both: the technology shift is structural, but the valuation excess riding on top of it is cyclical and mean-reverting. Getting that split right is the difference between a buying opportunity and a trap.

In July, Pearce argued that "in a bubble, everything rises, and we're not seeing that." That is the single most useful diagnostic in the debate. During the dot-com bubble, the entire technology complex re-rated together — the profitable and the unprofitable, the infrastructure and the application layer, the cash generators and the cash burners. What Pearce showed instead was a market that was discriminating with unusual precision. His chart split the four big hyperscalers — the companies spending hundreds of billions of dollars of free cash flow "chasing AI glory" — from the semiconductor companies receiving those cheques. Last financial year, the hyperscalers were down an average of 2 percent; the chipmakers were up an average of 138 percent. A bubble does not sort winners from losers with that kind of precision. A repricing does.

But that same July update located the froth elsewhere: in private markets. Pearce singled out OpenAI and Anthropic, the owners of the two leading AI models, as valuation concerns. That is consistent with how valuation excess actually forms — not in the liquid, scrutinized public companies, but in the unlisted assets where pricing is set by funding rounds rather than daily marks. The Firmus episode is the public-market version of the same disease: a company whose valuation depends on repeatedly tapping debt and equity markets at a time when the cost of capital has risen sharply.

The mechanism here is the cost of capital, not the technology. Ten-year U.S. Treasury yields sat at 5.15 percent on October 19, up from 4.67 percent in mid-July and 5.01 percent at the end of September. Pearce said he is comfortable with 5.5 percent yields, but that comfort is not free. At those levels, a business model that needs a lot more debt and more equity to fund expansion is not the same business model it was when rates were near zero. The AI super cycle is real on the demand side — Pearce lists it alongside excessive government spending as a driver of inflation — but the financing of that cycle is being repriced. The technology did not change. The price of money did.

The Second-Order Effect: AI's Real Transmission Channel Runs Through Bond Yields

The first-order story is familiar: hyperscalers are spending heavily on AI infrastructure. The second-order story is what that spending does to the rest of the market through the cost of capital. When the safest asset in the world pays above 5 percent, every risk asset must clear a higher hurdle. That does not kill AI. It sorts it.

Consider the free-cash-flow evidence Pearce raised in February. Alphabet's free cash flow, he said, ran to around $74 billion last year and fell to zero this year as spending shifted to AI. That is a dramatic number, and it cuts both ways. The bearish reading is that AI is consuming cash without producing it. The bullish reading — and the one Pearce's positioning implies — is that only companies with deep balance sheets and existing cash engines can afford to play. Alphabet can fund AI out of a search business that still prints cash; a data-centre developer cannot fund a $60 billion enterprise out of thin air.

This is why Pearce's fund remains structurally overweight U.S. technology while sitting out Firmus. The overweight is a bet on the companies receiving and deploying the capital, not on every company claiming a piece of the theme. It is also why he has said the fund would treat a 10 percent pullback in U.S. technology as a buying opportunity rather than an exit signal. The market, in his read, is anchoring financial assets to bond yields:

"We should be seeing a bit of turmoil in share markets because, after all, all financial assets are anchored to bond yields."

And yet equities have held. That resilience, with the 10-year yield above 5 percent, is the strongest evidence against a broad bubble.

The transmission channel runs in both directions, and that is what makes the current setup unusual. Higher yields compress multiples, which is the bearish leg. But higher yields also reflect an economy where government deficits and corporate capital expenditure are both running hot, which is the bullish leg. Pearce has it both ways: he names the AI investment super cycle as a source of inflationary pressure, and he still owns the companies doing the spending. The position is internally consistent only if the spending generates returns. If it does not, the two legs of the argument collapse into each other.

The counter-thesis is serious and deserves its due. The same July update that denied a broad bubble flagged concern about the data-centre investment cycle, and Pearce has warned that free cash flow previously used to increase dividends and buy back shares is now being used on "chasing AI glory." That diversion of cash is the classic late-cycle tell: companies that once returned capital to shareholders are now asking shareholders for more capital to fund growth. The hyperscalers are not issuing equity, but the economics are similar — free cash flow that supported buybacks now supports data centres, and the equity story depends on those data centres earning their keep.

If hyperscaler capital expenditure fails to generate returns — if the $74 billion of free cash flow does not come back, and instead keeps getting reinvested at low returns — then the equity correction stops being a valuation correction and becomes a price correction. Earnings expectations would fall alongside multiples, and the "valuation not price" comfort would evaporate. That is the scenario in which Pearce's no-bubble call fails, and it is not a fringe view. It is the bear case that every AI bull has to answer.

The falsifying signal is quantifiable. Pearce's call rests on the S&P 500 holding near its highs while yields stabilize. If the index falls more than 10 percent from its October 2026 peak while the 10-year Treasury yield remains above 5 percent, the market is pricing an earnings recession, not a multiple compression, and the no-bubble thesis is wrong. A second signal would be a genuine broadening of the decline: in a valuation correction, laggards hold up while leaders de-rate; in an earnings-driven correction, everything falls together — the very pattern Pearce said he is not seeing. A third would be the hyperscalers themselves: if their free cash flow does not recover as AI spending moderates, the chasing-AI-glory critique becomes an earnings story rather than a valuation story.

What Comes Next: Three Scenarios

The short-term outlook turns on the bond market. If 10-year yields stabilize around 5 to 5.5 percent — the range Pearce says he can live with — equity multiples can hold and the AI trade continues with stock selection doing the work. In that base case, the beneficiaries are the hyperscalers with cash generation, the semiconductor companies capturing the capital expenditure, and the software leaders that can layer AI onto existing distributions without eye-watering spending. The exposed are the capital-intensive infrastructure plays that must keep returning to markets, and the private-market AI names priced on funding-round momentum rather than cash flow.

The upside case is that AI spending begins to show up in productivity and earnings faster than expected, validating the capital expenditure and pulling the laggards along. In that world, the Firmus withdrawal is remembered as a one-off pricing failure rather than a regime change, and the companies that kept spending are rewarded for their conviction. The downside case is that sticky inflation forces traders to price in further rate increases, pushing the 10-year yield toward 6 percent — a level Pearce has flagged as capable of pausing the bull market — while AI monetization disappoints. In that scenario, the correction in valuation becomes a correction in price, and the Firmus withdrawal looks like the first of many rather than an isolated event.

For investors, the lesson is not to abandon AI. It is to recognize that the easy money in the theme has been made, and that the next leg belongs to companies that can fund their own ambition. Pearce's position — overweight the winners, sit out the priced-to-perfection deals, and demand that lenders stop financing weak AI business models — is a description of a market that is maturing, not cracking.

The sharpest takeaway is this: the AI boom is not dying; it is being underwritten. And in a 5 percent world, underwriting is a discipline that separates the builders from the storytellers.

Explore more exclusive insights at nextfin.ai.

Insights

Does Pearce fear an AI bubble burst?

What is the AI capex super cycle?

Why did Firmus shelve its IPO plan?

Why did Firmus valuation triple quickly?

What makes Firmus priced to perfection?

How does UniSuper view AI capex cycle?

How does UniSuper weight US tech?

Where does AI valuation froth exist?

Are private AI valuations too high?

Who owns leading AI models privately?

How do bond yields impact AI stocks?

How do yields anchor financial assets?

Are shares tied to bond market yields?

How high can Treasury yields rise?

What signals an AI earnings recession?

When does the AI bubble thesis fail?

How does free cash flow signal risk?

What limits Firmus expansion funding?

Can AI spending boost productivity soon?

What three scenarios face AI investors?

Search, ask or analyze anything...