UBS is telling clients that the artificial intelligence build-out still has room to run, even as the bill for it climbs into the trillions. In a note published this week, the Swiss bank’s chief investment officer, Ulrike Hoffmann-Burchardi, projected that global AI-related investment will climb from $1 trillion in 2026 to $1.5 trillion in 2028, with cumulative U.S. technology investment approaching $5 trillion over the same stretch.

That spending will not pay off evenly. Hoffmann-Burchardi cautioned that the industry will probably need to generate at least $1.5 trillion in revenue by 2030 to earn an adequate return, and that competition, price cuts, the cost of monetization and open-weight models all threaten those returns. External financing is getting harder to secure, she added, while market concentration and company-specific risks could amplify volatility.

The note lands at a jittery moment for the sector. Nvidia’s leadership troubles have weighed on sentiment, and investors are increasingly focused on whether the data centers, servers, cooling systems and power supplies required to sustain AI growth can be justified by future revenue. Demand for computing capacity, however, remains strong.

Against that backdrop, UBS analyst Karl Keirstead, whom the bank ranks among the top 3% of Wall Street analysts, has picked two established giants he believes can absorb rising costs and intensifying competition: Meta Platforms (META) and Amazon (AMZN). Both carry a Strong Buy consensus rating from the broader analyst community, according to the TipRanks database.

Meta: a cheap multiple, an expensive build-out

Meta, the $1.58 trillion parent of Facebook, Instagram, WhatsApp, Messenger and Threads, reported 3.64 billion daily active people across its apps in September 2026, up 3% from a year earlier. That reach underpins the advertising machine that funds its AI ambitions. Third-quarter revenue rose 26% year over year to $64.6 billion, beating forecasts by $1.2 billion, while adjusted earnings per share came in at $7.40, excluding a $9.89 billion litigation charge tied to a multistate legal settlement. Reported GAAP earnings were $3.83 a share.

The strain shows up in cash flow. Operating cash flow was a healthy $32.79 billion in the quarter, but capital spending of $34.9 billion pushed free cash flow to negative $2.12 billion, the company’s first quarterly cash burn in more than a decade, against positive $10.63 billion a year earlier. Meta has narrowed its full-year 2026 capex guidance to $135 billion to $145 billion and has announced a record $97.5 billion equity financing plan, including $25 billion of common stock priced at $625 a share and up to $60 billion of new non-voting Class C shares sold over time from early 2027. Nvidia and AMD have committed $3 billion and $1 billion as cornerstone investors.

Keirstead argues that enthusiasm for Meta’s new products can outweigh the balance-sheet strain. The company launched Muse in September, a personal AI agent that can run multi-step tasks in the background, from monitoring shopping prices to managing email and calendars, with a small-business version aimed at managing social media and finding customers. Meta also develops the Llama family of open-source models, the Meta AI assistant, virtual-reality headsets and AI-powered smart glasses.

“Positive investor sentiment about the prospect of Muse and Meta Business Agents may be enough to offset concerns about capex build, negative FCF and financing needs,” Keirstead wrote. “In our view the stock at 20x 2027 GAAP EPS seems attractively priced given the recent momentum in its AI model and apps (not just infra) story.”

The analyst rates the shares a Buy with an $815 price target, implying 32% upside over the next year. Meta closed at $616.38, down 12.6% over the past month, and the average analyst target of $857.24 points to a 39% gain. Of 43 analysts covering the stock, 37 rate it Buy and six Hold.

Amazon: the defensive AI play

Amazon, the world’s largest online retailer and a dot-com survivor, has seen its cloud unit become the engine of its growth. AWS generated $46 billion in revenue in the third quarter, up 39% year over year, far outpacing the 5% to 6% growth recorded by the company’s North America and International segments. Group net revenue rose 12% to $201.9 billion, roughly $340 million below expectations, while earnings per share of $1.89 compared with $1.95 a year earlier, a decline Amazon attributed largely to $9.5 billion in pre-tax gains from its Anthropic investments booked in the prior-year quarter.

Beyond renting out computing power, Amazon has moved into custom silicon. Its Trainium AI accelerator family has drawn commitments from model developers: Anthropic, the maker of Claude, has agreed to use up to 5 gigawatts of AWS Trainium capacity, while OpenAI has committed to about 2 gigawatts, with deployment ramping up in 2027. Amazon has disclosed that its chip business, including Trainium and its Graviton processors, is running at an annualized revenue rate above $25 billion.

Keirstead sees the breadth of AWS as a shield. “AI model and chip agnostic nature of AWS will help position Amazon shares as a defensive AI play and an AI enabler to enterprises scaling their AI investments in 2027+,” he wrote. “These positives outweigh the impact on AWS growth from quarterly variability in the model API access revenue stream.”

Even so, he trimmed his estimates, cutting AWS revenue growth for the fourth quarter of 2026 to 41% from 42% and for 2027 to 43% from 48%, while lowering his 2027 margin estimate to 40% from 45%. He rates the stock a Buy with a $340 price target, implying 27% upside. Amazon trades at $267.74, and the average target of $324.41 suggests a 21% gain. The stock carries a Strong Buy consensus, with 42 Buys and one Hold among 43 ratings.

The demand side

The bullish case rests on demand that shows few signs of fading. Amazon Chief Executive Andy Jassy said in a CNBC interview this week that the company’s large core computing capacity is essentially fully booked for 2027, with most of 2028 and a meaningful portion of 2029 also reserved. He described the growth of Amazon’s AI business in its first three years as roughly 440 times faster than AWS managed in its early years, and said AWS could exceed $1 trillion in annual revenue within five to 10 years.

Jassy also dismissed concerns that the boom is a bubble, pointing to a backlog of about $580 billion in which no single customer accounts for a quarter, and noting that more than 80% of Fortune 100 companies use its Bedrock inference service. He said he does not expect Anthropic’s initial public offering to change the partnership between the two companies.

MetricMeta Platforms (META)Amazon (AMZN)UBS rating / price targetBuy / $815Buy / $340Implied upside32%27%Recent share price$616.38$267.74Average analyst target$857.24$324.41Consensus (Buys / Holds)37 / 642 / 1Latest quarterly revenue$64.6 billion, +26% y/y$201.9 billion, +12% y/y

Note: Price targets and share prices as reported in the UBS note and TipRanks data.

What it means for investors

The UBS view captures a split emerging in the market. Companies with entrenched cash-generating franchises are being treated as safer ways to own the AI theme than the infrastructure suppliers whose spending plans carry more execution risk. Meta’s negative free cash flow and its planned equity raise are the clearest illustration of the trade-off: a stock trading at roughly 21 times trailing earnings looks inexpensive only if the AI spending eventually produces returns that justify the dilution.

For Amazon, the bet is on neutrality. By offering customers a choice of models and chips, including Nvidia’s GPUs alongside its own silicon, the company positions itself to collect fees regardless of which model developer wins. That diversification may offer some protection if, as Hoffmann-Burchardi warned, not every company reaps the rewards of the AI investment wave.