AI agents could make infrastructure that runs entire workloads more valuable than software layered on top of cloud platforms, according to two investors writing about Israel’s tech sector. Their argument is a useful challenge to the familiar idea that cheaper software creation simply means more software businesses.
Watch Desk analysis
What happened
Glenn Solomon and Dan Cahana of Notable Capital argue that agents writing and running more code will increase demand for the infrastructure beneath it. They point to GitHub recording nearly 1 billion commits last year and 2.9 billion in August this year, which they describe as a 35-fold jump in the monthly pace.
The authors say cloud-era software companies typically captured around 5–10% of the cloud spending associated with a workload, with the cloud provider taking the rest. They expect that share to shrink towards 1–2% as agents make workload-management software easier to replicate. By contrast, companies that run a workload themselves could capture more of the associated infrastructure spend. They call these businesses “workload clouds”, citing Snowflake, Vercel, fal and Browserbase as examples.
Why it matters
If the authors are right, the agent boom may favour companies that operate the underlying compute and services, rather than those selling another layer of management software. That is a consequential thesis for founders, investors and developers deciding where to build, though it is an investment argument, not settled evidence of where value will land.
The authors also identify security as a possible opening: agents can create environments, call services and act without a person reviewing each step. That makes securing agent activity a practical infrastructure challenge, not merely a question of making models more capable.
Our read
The strongest point is the shift in the unit of demand: agents could create and consume workloads at a pace that changes which parts of the stack capture the money. But the piece comes from investors making a case for a market they invest in, and its forecasts should be read accordingly. The useful test is whether real usage and margins support the “workload cloud” thesis, not whether the name sounds like it arrived with a pitch deck.
What to watch
- Whether agent-driven software production translates into sustained demand for new compute and infrastructure services.
- Whether workload-cloud businesses can grow while managing infrastructure costs and margins.
- Whether developers choose these services directly, and whether agent activity creates security needs that existing tools cannot meet.
Discussion spark: If agents make software easier to reproduce, will the lasting value sit with companies running workloads, or with the software built around them?
Sources and evidence
- Israel’s next AI giants could come from infrastructure | Ctech (16 October 2026, 08:38 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.