This Past Week was Huge for the AI Utility Theme
This past week was huge for the AI Utility theme.
The hyperscaler earnings all pointed to one thing that demand is running ahead of power, racks & chips which directly sets up the next leg for independent AI compute providers.
$Amazon.com(AMZN)$ showed AWS re-accelerating to 20% growth on a $132B run rate with backlog near $200B. They added 3.8 gigawatts of power in the last year and expect to double again by 2027. Every megawatt that isn’t delivered by AWS on time gets filled by third-party GPU infrastructure.
$Microsoft(MSFT)$ said AI capacity will grow 80% this year & their entire data center footprint will double in two years. They're capacity constrained through FY26 even after spending $35B on CapEx last quarter which means demand is spilling into non-hyperscaler compute providers that can deploy faster or cheaper.
$Alphabet(GOOG)$ Cloud backlog hit $155B, up 40% sequentially & management said they remain supply constrained through 2026. That’s an admission that the AI buildout is outpacing the speed at which power, silicon & data centers can come online.
$Meta Platforms, Inc.(META)$ raised CapEx to $71B for 2025 & said 2026 will be “notably larger.” They’re renting external GPU capacity while building their own which validates the economics of independent GPU providers who can offer compute to hyperscalers themselves.
This is the moment where specialized infrastructure players pick up share:
• $CoreWeave, Inc.(CRWV)$ high-density GPU cloud built for AI training at scale
• $Cipher Mining Inc.(CIFR)$ power-advantaged data center landlord for AI tenants
• $TeraWulf Inc.(WULF)$ baseload compute in the lowest-cost energy regions
• $APPLIED DIGITAL CORP(APLD)$ modular campuses deployed faster than traditional builds
• $IREN Ltd(IREN)$ lowest-cost renewable power driving cheapest cost per GPU hour
• $NEBIUS(NBIS)$ enterprise GPU clusters offering compute without hyperscaler lock-in
The next phase of AI infra belongs to whoever can deploy compute the fastest & most efficiently per megawatt.
Bar chart titled Exhibit 8 Change in capex estimates for AI hyperscalers showing all hyperscalers include AMZN GOOGL META MSFT ORCL. Vertical axis labeled in dollars from 100 to 800 billions. Horizontal axis with categories Start of 2025 Start of Q3 2025 Current. Blue bars for 2026 estimates orange bars for 2025 estimates. Start of 2025 shows 281 for 2025 and 400 for 2026. Start of Q3 2025 shows 314 for 2025 and 384 for 2026. Current shows 518 for 2025 and 400 for 2026. Annotations indicate 12 percent increase for 2025 at Start of 2025 58 percent increase for 2025 at Start of Q3 2025 19 percent increase for 2026 at Start of Q3 2025 and 29 percent decrease for 2026 at Current. Source FactSet Goldman Sachs Global Investment Research.
For whom haven't open CBA can know more from below:
🏦 Open a CBA today and enjoy privileges of up to SGD 20,000 in trading limit with 0 commission. Trade SG, HK, US stocks as well as ETFs unlimitedly!
Find out more here：
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
