ETO Markets｜Google Cloud Next'26 Outlook：OCS From Lab to Scale

Google Cloud Next26 will take place in Las Vegas from April 22 to 24.

According to Googles published agenda, the conference will feature an opening keynote on April 22, a developer keynote on April 23, and three days of sessions covering infrastructure, AI, developer technologies and industry solutions.

The central focus this year is whether OCS and silicon photonic networking will be formally elevated by Google as core pillars of next generation AI infrastructure.

Market Focus on Googles OCS Commercialization

Although Google has not disclosed specific OCS deployment plans ahead of Next26, industry direction is becoming increasingly clear. As AI clusters continue to scale, traditional electrical switching networks are approaching limits in bandwidth, power consumption and latency.

Optical switching is emerging as the key pathway for the next phase of AI infrastructure expansion. Within this framework, OCS and CPO have become the two most certain communication routes for AI accelerators.

Lumentum reported in its fiscal second quarter of 2026 that its OCS backlog has exceeded four hundred million dollars and that the company is rapidly expanding capacity to meet demand that is significantly above expectations. Coherent noted in its earnings call that OCS bookings continue to rise, with more than ten customer projects underway, and raised its multiyear serviceable market estimate to above two billion dollars.

Industry catalysts are also strengthening.

OFC 2026 positioned optical communication as a core theme, with emphasis on how data center scale OCS and photonic interconnects will support next generation AI compute architectures. Combined with the long standing focus of Cedric Lam and Google‘s technical infrastructure team on future data center networking and interconnect technologies, expectations for Google to reinforce OCS and silicon photonics at Next’26 are rising.

The foundation behind this momentum is Googles TPU demand entering a full scale expansion phase.

Reuters reported on April 6 that Broadcom has signed a long term agreement with Google to co develop and supply custom AI chips through 2031, directly supporting the next generation TPU rack architecture. At the same time, Anthropic announced a new agreement to procure up to three point five gigawatts of next generation TPU compute from Google starting in 2027. This signals that Googles TPU platform is no longer limited to internal use but is entering a cycle of large scale external delivery to model developers.

Liquid Cooling and Power

The Structural Constraints of AI Infrastructure

Beyond optical networking, liquid cooling and power systems represent another critical yet often overlooked theme of the conference.

Google has repeatedly emphasized that power availability and thermal capacity are the core constraints on AI data center expansion. In January, the company noted that the United States transmission grid has become a primary bottleneck for data center growth. In March, Google further disclosed that it is in discussions with suppliers, including Chinese vendors, regarding liquid cooling procurement.

This indicates that sustaining Googles next generation AI clusters requires more than TPU and OCS hardware. It also depends on power delivery, cooling and rack level infrastructure that form the secondary foundation of AI deployment.

ETO Markets Analyst View

Three Layers of Beneficiaries and Key Risks

From an investment perspective, the potential beneficiaries of Google Cloud Next26 can be divided into three layers:

- Core layer: TPU and ASIC design and supply chain

- Middle layer: OCS, CPO and high speed optical interconnect ecosystem

- Foundation layer: Power and liquid cooling systems supporting AI rack expansion

If the conference confirms accelerated Ironwood deployment and formally elevates OCS and silicon photonics within Googles cluster strategy, the Google investment narrative may shift from a narrow focus on Broadcom and TPU toward a system level theme covering ASICs, optical networking, liquid cooling and power infrastructure.

However, risks remain. If the conference places greater emphasis on software and enterprise AI applications while providing limited hardware disclosure, the markets pre priced expectations for OCS and the broader Google supply chain could face short term correction pressure.

Disclaimer

The information contained herein is for general reference only and does not constitute investment advice, a solicitation, or an offer to buy or sell any financial products.

ETO Markets does not guarantee the accuracy, completeness, or timeliness of the information and shall not be liable for any losses incurred from reliance on such content.