FEATURE: Global equities | Triaging tech's second act

"I think the odds are on our side."

That was the message from Anthropic chief executive Dario Amodei at the World Economic Forum in Davos, as the race for global dominance in artificial intelligence (AI) gains historic pace.

Anthropic is building the large language model Claude, designed to automate complex cognitive tasks and business workflows. It competes directly with OpenAI and Google to develop increasingly capable models, pushing the limits of reasoning, automation and machine cognition.

Yet building a powerful model is only one part of the equation. At Davos, Amodei spoke about a central dilemma: how much computing capacity must Anthropic lock in today to meet demand in 2027 - long before revenue visibility, customer adoption and pricing power are certain.

The fact that data centres - the backbone of AI computing - take one to two years to build, only makes Amodei's decision more consequential.

It is a narrow path. Buy too little compute and Anthropic may be unable to meet customer demand in 2027, effectively conceding ground in the race to its competitors.

Buy too much and it risks committing billions to capacity that revenue cannot support. In an extreme scenario, if demand for the models doesn't scale to expectation, Amodei spoke of the risk of going bankrupt.

Describing it as the "cone of uncertainty", he called the race to the top one filled with uncertainty.

"There is an inherent risk of overextension," he said.

The models built by OpenAI and Anthropic sit atop a broader ecosystem of companies supplying the chips, data centres and cloud infrastructure that power them.

Nvidia chief executive Jensen Huang describes the AI value chain as a five-layered cake. At its base is energy, the electricity required to power vast data centres. Above that sit chip designers and manufacturers, including Nvidia and Taiwan Semiconductor Manufacturing Company (TSMC), who design and produce the advanced processors that train and run AI models.

The next layer is infrastructure, dominated by hyperscale cloud platforms such as Amazon Web Services and Microsoft Azure, which provide the computing capacity, storage and networking required to deploy AI at scale.

On top of that sit the foundational model developers like OpenAI and Anthropic, and finally the application layer, where those models are used to build products and services tailored to specific industries and end users.

Not a simple cake, this one. Nor is it cheap to make. But the tech giants are ready for the splurge. Based on prior guidance issued, capital expenditure (capex) by the large tech cohort on AI infrastructure is expected to rise to as much as US$650 billion in 2026.

Individual commitments are even more striking. Alphabet plans to spend roughly US$175 - US$185 billion as it expands compute capacity, while Amazon recently committed close to US$200 billion in expenditure tied to infrastructure and cloud expansion.

Nvidia says AI capex could grow to as much as US$4 trillion by the end of the decade.

Everyone wants a slice of this cake. Capital from pension funds and global asset managers to retail investors is flowing into every layer, lifting valuations across the stack.

The tech sector in the US now accounts for more than a third of the S&P500's market capitalisation, around US$21 trillion in dollar terms. Nvidia alone has a market capitalisation of over US$4 trillion.

The scale of spending and the surge in valuations have raised concerns about the formation of a tech bubble as investors worry expectations may be running ahead of reality.

This is also not the first time markets have priced a technological revolution at speed. During the dot-com bubble in the early 2000s, valuations plunged and many companies once hailed as the future of the internet disappeared.

Some, such as Amazon and Microsoft, endured and went on to dominate. Yet the majority failed to justify the capital poured into them.

The uncertainty Amodei described remains central to the AI trade: will demand from the AI buildout ultimately justify the scale of the investments underway?

As money continues to flood into the sector, the next question also becomes unavoidable: what companies will turn out victorious in the race?

Pullback on pullback

Markets have endured extreme volatility in the past year. US President Donald Trump's trade policies, geopolitical tensions and shifting interest rate expectations have all unsettled sentiment, while the tech sector faces its own pressures.

The first pullback in the sector came in November last year as investors questioned the scale of capital being committed to AI data centres.

Munro Partners portfolio manager Kieran Moore says it was sentiment-driven, with investors worried that limited near-term revenue would not be sufficient to fund the surge in compute capacity being built over the next few years.

The second pullback came more recently in early February this year, when on average US software companies lost about 20% of their market value as of mid-month.

"It seems to me that every few weeks, a new AI tool is announced that threatens to revolutionise an industry," UniSuper chief investment officer John Pearce said in a recent investment update.

This time around, the sell-off followed concerns that Anthropic's latest model could encroach on traditional software businesses, with investors reassessing growth assumptions across the sector.

Pearce thinks the first big casualty of this tech revolution is the tech industry.

"There is an old adage that software is eating the world," he said.

"Now, it's AI's turn to eat software. That's of course before AI potentially eats itself. There is an irony here, isn't there?"

Opportunity and threats

Moore says the Melbourne-based investment manager doesn't think the trade is a bubble but a bull market opportunity that will endure for the next five to 10 years.

"We expect small DeepSeek-type moments and corrections along the way, but fundamentally we can see a path to some pretty meaningful upside for many of the beneficiary stocks over the next multi-year period," he adds.

China's DeepSeek rattled markets last year by being able to train its model at a fraction of the cost to its peers, casting doubts on the need for the infrastructure spend.

BlackRock Investment Institute chief investment strategist for APAC Ben Powell also sees the ongoing buildout as one of the defining forces shaping markets in 2026.

"This shift marks a move from a capital-light tech regime to a capital-intensive investment-led regime, creating opportunities for active investors to identify winners in the supply chain rather than relying solely on headline tech names," Powell says.

While the Magnificent 7 stocks will continue to play an important role, Powell notes there will be growing opportunities in the infrastructure behind AI including semiconductors, power systems and critical enablers.

Moore sees the power sector as an interesting bottleneck to find opportunities. He gives the example of Munro Partners' investment in Constellation Energy, which is a nuclear operator in the US with a long-term commitment to supply nuclear power to companies like Microsoft and Meta.

BlackRock is focused on 'picks and shovels' of AI where the spending is mostly front-loaded, demand is surging, constraints are likely to bite and barriers to entry are high. This includes players in chipmakers, grid operators, critical minerals and data centres.

ECP Asset Management principal, investments Annabelle Miller says the fund manager is being selective with its stocks, sticking to those it sees as long-term beneficiaries of deploying AI across product suites.

For example, ECP is increasing exposure to TSMC, calling it the "highest quality AI infrastructure beneficiary".

"TSMC is a pick due to its deepening moat and its monopoly on the advanced node technology required for all modern AI chips," Miller says.

"The company is very cautious about overbuilding capacity, and this supply tightness is driving prices up, allowing TSMC to command much higher margins from its customers, including Nvidia and AMD, as they guide toward a 50% compound annual growth rate for their AI semiconductors."

However, GQG Partners client portfolio manager David Jenkins says even though some existing businesses may be long-term AI winners, the fund manager views many of these businesses showing signs of deterioration.

"For us instead of picking a winner from a group of companies with declining earnings visibility, we believe the best way to compound capital is to focus on businesses where we feel earnings growth is more probable," Jenkins says.

Alvia Asset Partners is also treading carefully, as it sees a lot of capital chasing ideas bandied together under the AI thematic.

"We are overly cautious especially in the newer ventures into AI," Alvia portfolio manager Chris Scarpato says, adding that while Alvia embraces tech, valuations of some of the Magnificent 7 stocks are too "punchy" for them at the moment.

Alvia is backing established businesses adopting AI in their existing operations, where they have the free cash flow to make existing softwares effective for current users. Scarpato sees Booking Holdings and Constellation Software as some firms doing that.

Miller also sees ASX-listed Block as a firm unlocking AI efficiencies by automating workflows across engineering, customer service and operations. Block recently reduced its workforce by 40%, swapping human employees for AI.

However, even with the punchy valuations, State Street Investment Management senior strategist Clive Maguchu says the fundamentals of the Magnificent 7 are mostly very solid.

"They have generated a huge amount of free cash flows. They're very profitable businesses. They are well run. They don't have a huge amount of debt on their balance sheets yet," Maguchu says.

While he notes the free cash flow of Magnificent 7 stocks would just about cover the capex in 2026, he adds some firms will increasingly lean on external debt financing to supplement internal cash to continue the investment.

Miller agrees, highlighting Alphabet's recent issuance of a 100-year bond denominated in British pounds to fund its expenditures.

"But we do see strong investor support and demand for that kind of financing, because there is pretty good consensus around the value add that people are expecting from AI in the long term," he says.

Moore says when it comes to AI, the 'value add' needs to be thought of in more than just a "user times price equation".

"You've got to think about things like efficiency gains and productivity gains, time savings and customer satisfaction improvements," Moore says, noting globally every boardroom is talking about how they can deploy AI in their businesses to remain competitive.

He gives the example of Axon to showcase the productivity gains taking place in real time. Axon makes tasers and body cams for police officers to wear. Moore highlights how AI has started effectively taking the data during a police officer's shift and creating an automated report at the end of it.

"It's not a revenue generation opportunity for the police department, but it's an improvement in the ability to resource and allocate those resources," he says.

Jenkins, however, raises concerns over the revenue generation potential through the value chain in the trade.

Startups are the starting point.

"The AI startups have to be making money in order for the growth in cloud... that has to be real money for the growth to be real in order for the cloud providers to keep investing in the infrastructure," he says.

OpenAI and Anthropic are both unprofitable at the moment as they spend more than they earn to fund future growth.

Jenkins adds these startups lean on advertising for monetisation, which often draws from the same market share of incumbents that are investing in them.

"There's only so much advertising dollar to go around," he says.

And for the cloud side of the value chain, he says the quality and earnings growth of most businesses appears to be getting worse.

A new wave of neo-cloud providers optimised specifically to handle AI workload have entered markets, competing with hyperscalers that provide general-purpose cloud services along with catering to the AI market.

"Very simply, there's more competitors and the margins are declining; this isn't the same set up that cloud has seen in the past," Jenkins says.

Miller adds that lower input costs associated with running AI services will be important in driving return on investment. This includes expanding the supply of memory and power to drive down operating costs while improving the efficiency of custom silicon used in hyperscaler data centres, she says.

Scarpato also raises concerns on the circularity of the money being spent across the value chain forming a closed loop between investor, supplier and customer all at once.

"Nvidia might invest $100 million in a data center business, but then that data center company needs to buy Nvidia's chips, so it's just to basically prop up Nvidia's revenue," he says.

Jenkins mentions how companies will crack deals where neither the first company would have the actual compute capacity and the company buying it wouldn't have the money to pay for it.

"We've observed instances where without any money changing hands or potentially even the capacity to provide the services promised, revenue projections increase, and then their valuations go up. That then flows through the entire ecosystem," he adds.

Powell, however, says talk of a bubble is not a practical lens for investing as they tend to grow for some time and only become clear after bursting.

"Framing the debate that way also focuses only on the unprecedented level of spending, while ignoring the possibility that AI could be unprecedented in terms of the revenue it ultimately generates," he says.

While he notes that the buildout creates challenges around cash flows and returns in near term, he adds that investment plans are never on autopilot.

"They can adjust as constraints - especially around power and infrastructure - and revenue visibility evolve," he adds.

Pearce notes the current broadening outside the tech sector as healthy.

"You have to bear in mind one company's spending is another company's revenue, so everyone benefits," he says.

"That is indeed why we are seeing a broadening of the market rally outside tech. This is a very healthy development and once again demonstrates the benefits of diversification."

However, Scarpato notes that it won't take much for sentiments to turn sour considering how myopic and short-term the market has become.

"It won't take much and it's amazing how that capital can flow really quickly," he says.

He says Alvia will look out for any external shocks in macro environment which might impact flow of capital. Any pullback on spending by the big players could be another trigger.

Jenkins says it is closer than not for earnings to start to matter, and this can be seen in the sell-off over the last month. He adds the rate of spending should decelerate with businesses already exhausting free cash flow.

"The ramp up in capex has been incredibly fast, a lot faster than it was in the dot-com bubble.... there's no real dry powder left for [capex] growth to continue increasing," he says.

The higher risk portion of the market will be hit the hardest, Scarpato predicts, while the "big guys" will consolidate.

Race to glory

In its 2026 macroeconomic outlook, AustralianSuper said innovation waves almost always sponsor extraordinary gains in stocks - some of which prove durable and some don't.

"Innovations almost always bring benefits, but we don't know precisely when the benefits will come or how large the benefits will be," AustralianSuper said.

"We do know one thing, however: betting against innovation as an economic and market phenomenon is rarely, if ever, profitable over the long term."

Amodei agrees. With the inherent risk of overextension, he says it is important to separate the economic side from the technology itself.

"I am one of the most bullish people around [when it comes to AI]," he said.

And like Amodei trying to tip the economic odds in his favour, investors can only work to do the same.