Market Signals - Sunday August 2, 2026
Are stocks begining to price in the companies which will benefit from cheaper models and the agent economy?
What hyperscalers confirmed about 2026
We just finished a week of megacap earnings, and the hyperscalers are continuing their aggressive infrastructure spending plans for the rest of 2026.
Spending is still climbing steeply, none of them has reduced its underlying investment plan, and on their own guidance the four megacaps being tracked go from roughly $376B in 2025 to about $719B in 2026.
The 2027 visibility gap
The remaining question is no longer whether the 2026 buildout will happen. It is whether 2026 is the peak or a pause before another leg higher in 2027.
No hyperscaler has provided a formal or comprehensive 2027 capex plan, but GOOGL, AMZN and MSFT pointed past it during their earnings calls.
AI infrastructure beneficiaries will increasingly trade on evidence about 2027 demand, not merely execution against already-known 2026 budgets. Although analysts currently model substantial further capex growth at Alphabet, Amazon and Meta. Microsoft is the clearest illustration of the visibility gap, with no comparable 2027 figure yet available.
Who will fill this capacity?
The spending is real. And here is what the companies said about the demand they are seeing for their services.
"We continue to be supply constrained", per Pichai, and the CFO said it has now been true "multiple quarters in a row".
"Customer demand continues to exceed available supply", CFO Amy Hood of Microsoft.
Jassy, on the ~$220B capex plan: "Even at that amount, we will still not have enough capacity to meet all the demand we have in 2026, and I believe this dynamic will also be true in 2027, too".
The transition to cheaper models
On July 20th, a Chinese startup, Moonshot AI released the openweight Kimi K3 model with 2.8 trillion parameters, which competes directly with the leading models from frontier labs like Anthropic and OpenAI.
Openweight models significantly drop the cost of tokens and cheaper tokens make more AI applications economically viable. The companies with enormous installed customer bases who can now ship AI features to hundreds of millions of users at a fraction of last year’s cost.
Cheaper tokens increase the number of AI agents running in production, the volume of data moving across networks, and the amount of identities and systems enterprises must manage and secure.
This could create new bottlenecks:
Token infrastructure: More tokens require more compute, memory bandwidth, storage, networking and data movement. We know that Kimi K3 hit a GPU capacity wall within 48 hours of release and paused new subscriptions.
Agent management: More agents require orchestration, permissions, monitoring, governance and security.
I explored the fundamental implications in The Cheap-Model Winners: Who Cashes In When AI Gets Commoditized?
Is the Market Beginning to Price This Transition?
Let us look at a few charts.
Cybersecurity ETF CIBR 0.00%↑
As the number of agents increases securing the agents an enterprise’s network becomes an increasingly big problem.
Identity Management OKTA 0.00%↑
Okta is becoming the identity control plane for the agent economy, verifying each agent and governing what it can access and do. When asked about the opportunity in front of them during the Q1 call on May 28, 2026, Todd McKinnon the CEO said "The pipeline's bigger than anything we've ever seen."
Market is betting that OKTA will be one of the big winners of the transition to an agent economy.
Gitlab GTLB 0.00%↑
Gitlab’s Duo Platform enables it’s customers to orchestrate and govern AI agents within their enterprise. I talked about the technical setup earlier in a note.
Hewlett Packard Enterprise HPE 0.00%↑
HPE benefits from exploding token usage by supplying the compute and networking infrastructure needed to run and move inference workloads in both public and private clouds.
Nebius Group NBIS 0.00%↑
Nebius has built up the cloud capacity to globally run training and inference workloads. The stock retained all of it’s 27% gains from Thursday, and is sitting right at it’s 10/100 day. can it build a base from here?
Micron MU 0.00%↑
Micron is the bellwether US memory stock for SDRAM, HBM and NAND storage.
For now it found support at the 100 day in the 700+ area - can it build a base from here?
What happens next?
We are at an interesting spot in the market, and a lot depends on how the overall market behaves from here.
Is this a mean reversion in cybersecurity and software or is it a genuine rotation away from the semis? This is one of the key questions I am waiting to get answered by this market.
As we navigate this tricky market, it is important for the market to prove to us that it is ready for the next leg up. The goal is not to catch the exact bottom, rather it is to participate in the larger move while keeping the risk clearly defined.
Good luck!
Disclaimer: This article is for informational and educational purposes only and does not constitute financial, investment, tax, or legal advice. Any opinions, scenarios, price targets, or market observations reflect my personal views and may change without notice. Investing and trading involve substantial risk, including the possible loss of principal. You are solely responsible for your own investment decisions, position sizing, risk management, and trades. Conduct your own research and consult a qualified professional where appropriate.