UBS's 8,100 Call Hides an $820B Power Bill Nobody's Pricing In

UBS just raised its year-end S&P 500 target to 8,100, up from 7,500, and the number that should grab investors' attention isn't the index level — it's the $820 billion the bank says is about to get spent building the infrastructure behind it.

Photo by Rafael Minguet Delgado on Pexels

What Happened

UBS lifted its 2026 year-end S&P 500 target to 8,100 from 7,500, according to a note reported by CNBC on July 21. With the index trading around 7,440 at the time, the new target implies more than 8% upside over the back half of the year. UBS strategists pointed to earnings power: they expect S&P 500 profit growth of more than 28% this year, above consensus, with a large share of that driven by semiconductor companies working through mounting order backlogs. The bank also forecasts AI-related capital expenditure jumping 68% year-over-year in 2026 to roughly $820 billion, followed by another sizable increase in 2027 toward the $1 trillion mark that other Wall Street desks, including Evercore and Bank of America, are now floating for hyperscaler spending.

UBS's 8,100 target now sits as the second-highest on the Street, trailing only Oppenheimer's 8,150 call. Citi set an identical 8,100 target back in early June, meaning three major banks have now converged on roughly the same number within weeks of each other — a level of agreement that's notable given how differently these firms have modeled the AI trade over the past year.

Why It Matters: The Capex Number Is the Real Story

As I laid out in Behind UBS's 8,100 Call: Chips Do Half the Heavy Lifting, the semiconductor earnings engine is well understood at this point — memory pricing, backlog conversion, the Micron and SK Hynix side of the trade. What's less discussed is where that $820 billion actually goes once it leaves a hyperscaler's balance sheet. A large chunk of AI capital spending isn't chips at all — it's power generation, grid connections, cooling systems, and the physical real estate to house it all. Global data-center electricity demand is on pace to roughly double from about 485 terawatt-hours in 2025 to near 950 terawatt-hours by 2030, and AI-optimized workloads already account for roughly a third of the industry's power draw. That buildout doesn't show up in a chip earnings report — it shows up in utility capex plans, grid interconnection queues, and turbine order books.

Photo by QuinceCreative on Pixabay

Who's Affected

The equipment and power side of the trade has already started moving on hard numbers, not just sentiment. Vertiv, which makes data-center power and thermal management systems, reported Q1 2026 sales of $2.65 billion, up 30% year-over-year, with adjusted operating profit up 64% and Americas revenue growing 44% organically — strong enough that the company raised full-year guidance to $13.5–$14.0 billion in net sales. GE Vernova posted Q1 2026 orders of $18.3 billion, up 71% organically, including $2.4 billion in electrification equipment orders tied to data centers in a single quarter — more than it booked in all of 2025 combined. On the generation side, Constellation Energy became the largest electricity producer in the U.S. after closing its $16.4 billion Calpine acquisition in January, and its restarted Three Mile Island unit, now called Crane, is already under a roughly 835-megawatt supply contract with Microsoft. None of this is semiconductor demand — it's the physical grid straining to keep up with it.

• Semiconductors: Still the earnings backbone of UBS's 28% growth call, as covered previously on this blog.
• Power equipment and utilities: Vertiv, GE Vernova, and Constellation Energy are now booking orders directly tied to data-center buildout, not general industrial demand.
• Hyperscalers: Microsoft, Amazon, Alphabet, and Meta are the ones writing the checks — their capex guidance is the leading indicator for whether the $820 billion figure holds up.

What to Watch Next

The next real test isn't UBS's index math — it's whether hyperscaler capex guidance in upcoming earnings reports actually confirms the $820 billion trajectory, or whether any single company signals a pullback. A slowdown in cloud capex commentary from Microsoft, Amazon, Alphabet, or Meta would ripple through the power and equipment names faster than it would hit chipmakers, since utilities and power-equipment makers are further downstream and more exposed to project timing. Watch order backlogs at Vertiv and GE Vernova in their next quarterly updates, and watch grid interconnection and power-purchase-agreement announcements from utilities — those are the clearest real-time signal of whether the capex bet is accelerating or stalling.

This is not financial advice — always do your own research before making investment decisions.

UBS's 8,100 target is built on an earnings and capex story that spans far more than chips — it now includes a physical infrastructure buildout that's already showing up in equipment backlogs and utility earnings. Whether that trade keeps working depends less on where the S&P 500 index sits in December and more on whether the money hyperscalers have promised to spend actually gets spent on schedule.