Welcome to Issue 162 of The CTO Show Brief.
Agents Are Now Workers. The Infrastructure Bill Comes Due.
Three things converged this week: AI moved from assistant to executor, the physical substrate powering that shift hit a genuine capacity wall, and the capital allocating to close the gap became unmistakably sovereign. The speculative era ended quietly. What replaced it is industrial, expensive, and harder to replicate than software.
Gigawatt-Scale Is the New Unit of Competition
Broadcom shipped the 102.4 Tbps Tomahawk 6 switching chip and unveiled a 400G/lane optical DSP this week. These are not speed records. They are the minimum specs required to run clusters exceeding one million accelerators. Simultaneously, Nscale closed $2B at a $14.6B valuation to build out AI infrastructure at hyperscaler density.
The constraint is no longer silicon. It is energy. AI data centers need an additional 92 gigawatts by 2027, and turbines for gas generation are already sold out. Power availability is now a primary investment risk metric alongside team and product.
Who gains leverage: Operators with behind-the-meter generation or locked energy agreements. Anyone holding 'power permits' as a moat. Startups building power-efficient silicon architectures in SiC and GaN.
Microsoft Redefined the Enterprise Worker. It Is Not Human.
Microsoft 365 E7, the Frontier Suite, launched at $99/user/month. The first new enterprise tier since E5 in 2015. The product bundles E5, Copilot, Agent 365, and the Entra Suite. Its core function is governing AI agents, assigning them identities, access controls, compliance scopes, and audit trails.
This is the institutionalization of digital labor. Microsoft is betting that by 2027, the distinction between 'employee seat' and 'agent seat' collapses into a single governance problem. The 8% price increase on E3 and E5 simultaneously makes E7 the mathematically forced upgrade path for Copilot-invested organizations.
Who gains leverage: Vendors that integrate cleanly with Agent 365's identity model. Startups building agent-specific IAM and observability tooling. Those that don't are now platform risk.
Agentic Security Became a Market Category Overnight
OpenAI acquired Promptfoo, an AI red-teaming platform used by over 25% of the Fortune 500. The same week, Onyx Security launched with $40M to build a 'secure AI control plane' that monitors and blocks agent actions in real time. These moves are not defensive. They are preconditions for enterprise deployment.
The risk profile of autonomous agents is categorically different from chat AI. Indirect prompt injection, agent-to-agent exploits, and out-of-policy execution are functional risks, not information risks. An agent that can file invoices, trigger purchases, or modify configurations requires a governance layer before production, not after. The EU AI Act's August 2026 enforcement deadline is accelerating the timeline.
Who gains leverage: AI-native SOC platforms, traceability-as-a-service providers, and any vendor that can offer audit trails for agentic workflows. Bare AI platforms without embedded security are losing enterprise deals now.
Gulf Sovereign Capital Is Moving from Buyer to Builder
Saudi Arabia's PIF is restructuring its 2026-2030 roadmap away from real estate mega-projects toward AI infrastructure, industrial localization, and strategic minerals. Abu Dhabi's MGX, alongside BlackRock, is anchoring a $30B AI infrastructure fund. Gulf sovereign funds collectively deployed $126B via state-owned investors in 2025, 43% of global capital in that category.
The shift is structural, not cyclical. Fiscal pressure from sustained low oil prices is forcing a prioritization of durable leverage: control of AI infrastructure, the mineral supply chain for chips, and the energy required to run both. The PIF is transitioning from passive LP to infrastructure operator. Founders targeting MENA should price that accordingly.
Who gains leverage: Startups offering AI data centers at the edge, hybrid infrastructure, and anything at the intersection of mining technology and green energy. The region is building vertical integration from raw materials to compute.
The $1B Seed Round Is a Signal About Scientific Differentiation, Not Valuation Excess
Advanced Machine Intelligence (AMI), associated with Yann LeCun, raised over $1B in seed funding at a $3.5B valuation. Thinking Machines Lab (Mira Murati) secured a strategic partnership with Nvidia that provides guaranteed access to next-generation accelerators. These are not hype rounds. They reflect a thesis: capital is moving back toward fundamental platform bets, not application interfaces.
The VC market is bifurcating cleanly. Elite technical founders with novel architectural propositions attract uncapped seed-stage capital. Everyone else faces compressed valuations and stricter unit economics scrutiny. The AMI deal also signals Europe is forming AI platforms capable of competing at foundational model depth, not just application layer. Secondary markets have exceeded $210B as primary exits stall.
Who gains leverage: Founders with genuine scientific differentiation and a path to embodied or physical-world AI. Investors with the patient capital horizon to back long-cycle research bets. Application-layer AI without proprietary data or architecture is compressed.
The forward question: When AI agents hold enterprise identities, access controls, and audit trails, the liability question becomes unavoidable. Who is legally responsible when an agent executes incorrectly? That answer will determine the next wave of compliance infrastructure.
🎙️Episodes Recap:
Enterprise transformation rarely fails because of strategy. It fails because of execution, and one of the most complex parts of execution is data migration. In this episode, Mehmet speaks with Dominik Wittenbeck , CTO at SNP, about the real mechanics behind SAP transformations and why data migration is often the most underestimated phase of enterprise modernization. They explore how organizations approach SAP migrations, the risks of underestimating data transformation projects, and why Selective Data Transition (SDT / Bluefield®) is becoming a preferred strategy for many enterprises. The conversation also dives into how AI is beginning to reshape large-scale IT transformation projects, from presales and planning to testing and root cause analysis. Dominik shares practical insights on how AI can augment consultants rather than replace them, helping organizations manage increasingly complex system transformations with greater speed and accuracy.
As AI rapidly reshapes how software is built, technology leaders face a growing tension between speed and security. Development cycles are accelerating thanks to generative AI tools, while cybersecurity teams are struggling to keep pace with new risks introduced by AI-generated code, autonomous agents, and evolving cloud architectures. In this episode, Mehmet speaks with Ben Wilcox , CTO and CISO at ProArch, about how modern technology leaders balance innovation with risk management. The conversation explores the convergence of engineering and security leadership, the maturity gap in DevSecOps adoption, the implications of AI-assisted development, and the governance challenges organizations must address as AI becomes embedded in enterprise applications. Ben also shares insights on secure-by-design engineering practices, the changing role of CTOs, and why AI governance and visibility will become critical priorities in the near future.
📖 From Nowhere to Next
Every week I share startup lessons and stories through The CTO Show Brief. But if you want to go deeper, my book From Nowhere to Next brings together the experiences and insights that shaped my own journey.
Thanks for reading — and for being part of this growing, global-minded network.
— Mehmet
