Do you think compute from AWS/Azure/GCP will get so cheap that Adobe and ServiceNow pricing power will remain even if they needed a lot of compute and API's through Firefly/NowAssist?

REDDIT.COMApr 20, 1:32 AM UTC

Key insights

  • The article discusses the deflationary nature of compute and its impact on software companies like Adobe and ServiceNow. While compute costs decrease, these companies are expected to maintain subscription prices, leading to margin expansion. The high CAPEX and energy requirements for data centers create a barrier to entry, favoring large players like Amazon and Microsoft. This dynamic could positively influence US equities by supporting the profitability of software companies.
Do you think compute from AWS/Azure/GCP will get so cheap that Adobe and ServiceNow pricing power will remain even if they needed a lot of compute and API's through Firefly/NowAssist?

The Compute Paradox: Is Big Tech Building a Moat or a Commodity Trap?

​Does the trajectory of compute pricing change your Big Tech DCF models, or are you just betting on the margin expansion of software users like Adobe and ServiceNow?

​1. Compute is Inherently Deflationary

​While Moore's Law is slowing, Huang’s Law (exponential GPU growth) has taken over.

​Hardware Efficiency: New architectures (Blackwell/Rubin) provide massive "compute per watt" gains. The unit cost of a single AI token is crashing.

​The ASIC Shift: Custom chips (AWS Trainium, Google TPU) are "stripped-down" for AI. These offer inference at a 30-50% discount compared to general-purpose GPUs.

​2. The Margin Expansion Play

​In a market with five major players (AWS, Azure, GCP, Oracle, CoreWeave), basic compute (IaaS) is becoming a commodity.

​API Price War: OpenAI/Azure API costs have dropped 90% in 18 months.

​The Software Win: As unit costs drop, companies like Adobe and ServiceNow will likely keep their subscription prices stable, pocketing the savings as pure margin expansion. Even if their total compute spend rises due to volume, their efficiency scales faster.

​3. The New Moats: CAPEX and Energy

​If compute is getting cheaper, why aren't there more competitors?

​The $100B Barrier: The entry price for a top-tier data center is now $50B–$100B in annual CAPEX. Very few companies on earth can play this game.

​Power is the Bottleneck: The constraint is no longer the chip; it’s electricity. Providers securing nuclear and renewable contracts (Amazon/Microsoft) have a cost advantage that silicon can’t beat

Continue reading on REDDIT.COM

Related Articles