From 30 GWh grid batteries to quantum-safe locks

Biggest battery by energy: Google and Xcel announced plans for a 30 GWh iron-air battery, expected to be the world’s largest grid battery by energy capacity.

Multi-day duration: The system comes from Form Energy and is meant to discharge for around 100 hours, covering multi-day periods of low wind and sun.

AI-driven demand: The project will help power a new Google data center backed by more than 1.5 GW of renewables in the region.

Follow the power markets: Returns will depend as much on tariffs and regulation as on chemistry. Some regions will make long-duration storage bankable; others won’t.

New contract structures: Hyperscalers are effectively becoming anchor tenants for bespoke grid assets, blurring lines between corporate PPAs and classic project finance.

Think stranded peakers: If multi-day storage becomes financeable, some gas peaker plants move from “core asset” to “back-up insurance.”

Hardware concentration risk: When diligencing AI funds, ask how they model dependence on one GPU vendor.

Public co-investment: Governments are subsidising validation; private capital still has to underwrite scale-up and go-to-market.

Listed exposure: Today’s clearer plays are in semis, IP vendors, design tools and test equipment; ARIA-backed work is early-stage and private.

Attention-grabbing claim: Researchers at the Advanced Quantum Technologies Institute released a paper describing the JVG algorithm, suggesting a faster path to cracking today’s public-key crypto.

Immediate pushback: Cryptographers highlighted major gaps: ignoring state preparation costs, confusing ideal qubits with real hardware, and testing only tiny toy problems.

Quiet progress on defenses: FIU researchers built and tested a quantum-safe encryption scheme for video, which slightly outperformed advanced classical methods on speed and security in lab tests.

Timelines can shift suddenly: Even if JVG itself falls apart, it shows that algorithmic tricks, not just more qubits, can compress risk timelines.

Harvest-now, decrypt-later: Sensitive data recorded today could be decrypted years from now if adversaries store it and quantum attacks get better. Long-lived data is already at risk.

Defensive tools are maturing: Quantum-safe schemes are moving out of theory into real-time applications like video, giving network and cloud providers something concrete to deploy.

Board-level question: Not “Is this one algorithm real?” but “How quickly can we swap out our crypto?” Many large systems need 7–10 years to fully upgrade.

New industry layer: Expect a growing post-quantum migration business: consulting, hardware modules, key management and secure protocols wrapped as services.

Look for turnkey offerings: Value may accrue to infrastructure and cloud providers who can offer one-click quantum-safe modes, not just whitepapers.

From prediction to generation: The team built a generative model for biology, moving beyond “How does this protein fold?” to “What could we build to get this function?”

HPC-driven search: The system runs on large compute clusters, exploring many possible designs in silico before any DNA is synthesized.

Broad ambitions: Potential applications range from new materials and enzymes to entirely new “chassis” organisms for industrial use, though those remain hypothetical for now.

New design power: This shifts biotech closer to chip design: simulate thousands of options, then physically build only the best ones.

Bigger biosecurity surface: The same tools that design useful microbes could, in principle, design harmful ones. Guardrails and access control become as important as accuracy.

Convergence story: Synthetic biology, AI and cloud compute are merging into an integrated design-build-test pipeline, which could compress development timelines dramatically.

Ask about the loop: In any synthetic-bio or platform company, key question is how AI plugs into their lab workflow and how results are measured.

New bottlenecks: DNA synthesis, automated labs and safety review committees may become harder to scale than model training.

Platform, not point solutions: Long-term value likely accrues to players who combine models, lab capacity, and a credible safety framework, rather than those owning a single tool.

Sierra’s mega-round: Sierra Space added $550 million in new capital, bringing total funding since 2021 above $2 billion, and framed itself explicitly as a defense-tech space company.

European launch build-out: PLD Space raised €180 million to industrialize its small launcher, Miura-5, and is planning a major new factory site in Elche, Spain.

Industrial-base narrative: Both companies emphasize regional jobs, supply chains and security, aligning their story with government industrial-policy goals.

Space as infrastructure, not novelty: Investors are betting on secure communications, earth observation and military payloads, not tourism.

Sovereign access: Europe is trying to plug the gap left by lost Russian launch services with home-grown rockets, which can shape which constellations and services survive.

Policy-anchored revenue: Government frameworks and defense contracts can give these firms longer-dated, stickier demand than purely commercial launch providers.

Different risk buckets: Upstream launch hardware, in-orbit services and data analytics each have distinct capex, regulation and exit paths.

Backlog quality matters: Look at contract length, cancellation clauses and re-bid risk, not just headline valuation.

Downstream leverage: Launch valuations implicitly assume a pipeline of satellites and services; due diligence should cover that downstream stack as well.

Paradigm broadens mandate: After years as a top crypto specialist with more than $12.6 billion under management, Paradigm is raising a new fund – up to $1.5 billion – that can back AI, robotics and broader frontier tech alongside crypto. 

Rationale for the shift: The firm sees rising overlap between crypto and AI, including agentic payments where AI systems transact autonomously, and wants flexibility to pursue the full stack. 

Tesla narrative pivots to robots: Tesla’s EV revenue fell in 2025 and net income dropped about 46%, but the stock is buoyed by expectations around Optimus robots and autonomous driving, with some forecasts pointing to multi-trillion-dollar valuations by 2027 if those bets land.  

Capital follows the full stack: Paradigm’s move shows that “frontier tech” funds now expect to cover compute, models, hardware robots and underlying financial rails in one vehicle, rather than siloing crypto. 

Physical AI story goes mainstream: Tesla is repositioning itself from an automaker to a robotics and AI platform, with public-market investors increasingly willing to value software and robots ahead of realized cash flows. 

Humanoids as asset class, not gadget: If Optimus or competitors reach meaningful deployment in factories and services, humanoids shift from “cool demo” to long-life productive assets – with capex, depreciation and financing structures to match.

Check mandate drift: For LPs in “crypto” funds, this is a moment to revisit fund documents and exposure maps – what exactly counts as core, and how does that change risk?

Separate narrative from evidence: Tesla’s humanoid and robotaxi stories are still largely pre-revenue and pre-scale; useful questions are about trial hours, safety record and unit economics rather than price targets. 

Watch second-order deal flow: Expect more crossover deals where AI, robotics, crypto rails and industrial automation show up in the same cap table – making ecosystem maps more important than single-company views.