TL;DR — Ahead of Alphabet’s February 4 earnings, three operating facts define the AI build-out: power contracts now gate data center capacity, Project Genie moves generative AI into production work, and AI glasses target the screen after the phone. This piece maps each one with data rather than forecasts.
As Alphabet approaches its February 4 reporting date, the useful questions are physical ones. Where does the electricity come from, what do the models produce beyond answers, and what device carries them next? Rather than forecast the quarter, this note lays out the operating facts already on the record.
The starting point is scale. In its third-quarter 2025 release, Alphabet raised its 2025 capital expenditure guidance to $91 billion to $93 billion and reported a Google Cloud backlog of $155 billion. Spending of that size only makes sense if the inputs behind it, power first, can be secured.
Power Contracts Now Gate Alphabet’s AI Capacity
AI scaling is increasingly bottlenecked by power procurement, not just chip supply. A data center full of accelerators is useless without a grid connection and firm generation behind it. Alphabet has started treating electricity as a long-term supply contract rather than a utility bill.
The clearest example is NextEra Energy (NEE). In October 2025, Google signed a 25-year power purchase agreement that supports the restart of NextEra’s Duane Arnold nuclear plant in Iowa. On December 8, 2025, the two companies expanded the partnership to jointly develop multiple gigawatt-scale data center campuses with their own generation, starting with three sites already in development.
These agreements work as the foundation layer for AI capacity. Long-dated power contracts lock in supply before rivals can, and they make energy reliability a prerequisite for computational growth rather than an afterthought. Chips can be ordered in quarters. New generation takes years.
Project Genie Pushes Gemini From Answers Into Production
On January 29, 2026, Google began rolling out Project Genie, a world model that generates interactive 3D environments from a text prompt or image, to subscribers of its top AI Ultra tier in the US. It marks a shift from AI as a consumer tool to AI as a production engine. Video game development is the test case because it is expensive, slow and labor-heavy.
The market treated it as a threat. According to Bloomberg, on January 30 Unity fell 24%, Roblox dropped 13% and Take-Two lost 7.9%. That reaction is a real-time read on how investors price the disruptive potential of generative tools against traditional development models.
The product itself is early. Generations are short sessions at modest resolution, with no persistent game world. The stock moves priced a possibility, not current revenue, and that distinction matters for anyone extrapolating from one trading day.
AI Glasses Are Alphabet’s Bid for the Screen After the Phone
The hardware focus is moving toward AI-powered glasses. On December 8, 2025, Google said the first glasses built on Android XR and Gemini would reach the market in 2026, made with Samsung, Warby Parker and Gentle Monster. Warby Parker had earlier disclosed a commitment from Google of up to $150 million.
Google described two categories: screen-free glasses with speakers, microphones and a camera, and a version with a small in-lens display. The objective is to move AI from a reactive interface, a search box, to a proactive one worn all day. If Gemini becomes the operating layer of wearable hardware, Alphabet captures the next screen before competitors do.
Taken together, the case for Alphabet heading into the print rests on three fixed variables:
- Secured energy for the data center fleet, through long-term contracts such as the NextEra partnership;
- Automated production, with Genie as the first public test of world models in a commercial category;
- Wearable AI hardware, with partners that bring retail reach and frame design Google lacks.
Power, production and the next screen: these are the fixed variables in the $GOOGL equation heading into February.
Positioning into the report reflects a consensus on core search resilience and on the monetization of cloud AI infrastructure. The chips that sit inside these data centers are covered in our TPU hardware analysis.
Update: The Q4 Print Put a Price on the Build-Out
Update, October 2026. The February 4 release confirmed the direction and raised the stakes. Q4 2025 revenue reached $113.8 billion, Google Cloud grew 48% to $17.7 billion and its backlog rose to $240 billion. Management guided 2026 capex to $175 billion to $185 billion, raised it to $180 billion to $190 billion in April and to $195 billion to $205 billion in July.
The cost of the build-out is now visible in cash flow. Second-quarter 2026 free cash flow was negative $5.9 billion on $44.9 billion of quarterly capex, even as Google Cloud’s operating margin rose to 35.6%. Our Google Cloud analysis tracks whether that margin can carry the spending.
The three pillars have held up as operating stories. What changed is the funding burden: Alphabet issued debt in each of the first two quarters of 2026 and raised equity in June, so the build-out is no longer paid for out of free cash flow alone.
What to watch next: new long-term power agreements and their timelines, wider availability and pricing of Genie, the launch date and price of the first Android XR glasses, and each quarter’s capex against operating cash flow. The energy pillar is the one with the longest lead time, so delays there would matter most.
Sources: Alphabet earnings releases (Q3 2025, Q4 2025) and 10-K and 10-Q filings on SEC EDGAR; quarterly materials at Alphabet Investor Relations; NextEra and Google Cloud joint announcement of December 8, 2025; gaming stock moves via Bloomberg. Disclosure: the author is long Alphabet (GOOGL/GOOG). Nothing here is investment advice; see the About page for our full disclosure.
