Recorded Jan 16th 2026. Exponential demand growth and AI mean the 2026 conversation is shifting from simply ‘power and data centres' to who can actually build what and when, as clean, reliable electrons grow scarce.

We track this shift through the week’s deal headlines, which naturally fall into layers of the same system. Read in sequence, they show how electricity constraint propagates upward from rack-level efficiency and inference per MW, to securing alternative compute architectures, vertically integrating generation ownership, procuring firm nuclear & geothermal power, and ultimately orchestrating an increasingly complex grid.

Incremental efficiency gains still matter, but as grid constraints tighten, the system increasingly needs technologies that materially change power requirements, deployment flexibility, operating envelopes, and siting options.

1. Groq’s $20 billion non-exclusive licensing agreement with Nvidia reframes chip efficiency as grid strategy: if you can do more inference per megawatt, you can deploy AI where others can’t, as power-efficient inference translates directly into deployable capacity when power delivery to a rack is constrained.

2. OpenAI’s multibillion-dollar, multi-year commitment to Cerebras goes beyond marginal optimisation by securing alternative compute architectures at scale with distinct power and siting characteristics.

3. Google/Alphabet’s $4.75 billion acquisition of Intersect Power shows that when efficiency gains aren’t enough, hyperscalers move to vertically integrate and own the power pipeline, as uncommitted, secure, grid-ready capacity and optionality become the most valuable strategic assets — reflecting the reality of interconnection queues and time-to-power.

Control over power supply then raises the question of quality. Intermittent renewables and batteries are improving, but hyperscalers are also looking for long-duration, clean firm powerm not from a decarbonisation perspective, but for reliability, security, and scale.

4. Meta signed three nuclear deals for up to 6.6 GW with Vistra, TerraPower, and Oklo.

5. Fervo Energy raised a $462M Series E, positioning enhanced geothermal as a complementary route to 24/7 clean power that could scale faster and across more geographies than nuclear alone.

6. Octopus Energy’s $1 billion raise to spin out Kraken Technologies at an $8.65 billion valuation shows that as the system becomes more distributed and complex, coordination and grid orchestration are becoming as critical as generation itself, increasingly determining who gets power, when, and how efficiently.

We then bring it home to the UK, unpacking the latest offshore wind auction outcomes and what they signal, the north–south transmission bottlenecks driving curtailment and balancing costs, and why “cheap renewables” don’t automatically translate into cheap bills without major grid buildout.

We explore why prices can go negative while gas still runs, why strike prices can rise even as renewables mature, and why the UK continues to avoid locational pricing—for now. The energy trilemma remains real, but in today’s environment, resilience and security are increasingly setting the agenda.