Nvidia–Groq’s $20B AI chip deal, record space funding, humanoids to $200B and Europe’s fusion push.

Groq and Nvidia signed a non-exclusive licensing agreement for Groq’s AI inference technology, with the aim of accelerating inference at global scale.

Groq’s founder and CEO Jonathan Ross and other senior engineers are moving to Nvidia, effectively making this a technology + talent transaction.

The deal value is reported at roughly $20B, characterised as Nvidia’s largest transaction and a major bet on inference-optimised architectures.

Confirms that the key bottleneck has shifted from training to inference capacity and latency; Nvidia is willing to spend tens of billions to capture that layer.

The non-exclusive licence + acquihire structure appears designed to limit antitrust exposure while still absorbing a challenger’s IP and talent into Nvidia’s ecosystem.

It signals that alternative architectures (like Groq’s SRAM-heavy, deterministic inference design) are strategically threatening enough to warrant defensive mega-deals.

Treat advanced AI chips as strategic infrastructure, with exits that may look more like complex licensing/partnership hybrids than clean M&A.

Ask managers how they model hardware concentration risk when the leading vendor can write $20B cheques to neutralise emerging competitors.

For private portfolios, benchmark AI-hardware theses against this deal: what is the realistic exit universe for ASIC and accelerator startups in a Nvidia-centric world?

Barclays lays out base and high-case projections for humanoid robotics, with an optimistic scenario approaching a $200B market.

It identifies three primary exposure routes: component makers (actuators, motors, sensors, semis), defence/robotics synergies, and rare-earth/critical minerals.

The analysis references direct humanoid and “physical AI” developers, many private or pre-profit (Tesla’s Optimus, Boston Dynamics, Figure, 1X, others).

This is one of the clearest mainstream attempts to size humanoids as a discrete theme, not just an offshoot of industrial automation.

Emphasis is on value accruing up the stack – bearings, actuators, chips and minerals – before general-purpose humanoid OEMs reach scale and margins.

The wide scenario band ($40–200B) underlines high uncertainty across technology, regulation, labour markets and safety frameworks.

In public markets, focus on today’s cash-flowing exposure (industrial robotics, components, mining) vs long-dated humanoid optionality priced into stories.

Ask managers to surface assumptions on labour substitution, liability and safety rules – these will shape adoption curves as much as hardware.

For private allocations, scrutinise the path from demo to deployment in warehouses, logistics and factories; impressive prototypes don’t guarantee unit economics.

At a European Parliament hearing, fusion companies and supporters argued that practical fusion power could be possible in 5–10 years, making now the time to build an industrial base.

The Fusion Industry Association and others are pushing for fusion-specific rules, separating it from fission in regulation and safety frameworks, and for milestone-based funding.

Europe currently trails in private fusion capital, with less than 5% of global fusion funding going to EU startups, according to industry figures cited at the hearing.

Marks a narrative shift from “fusion as science” to “fusion as bankable infrastructure”, with Brussels now explicitly discussing industrialisation pathways.

Fusion-specific regulation and funding instruments will heavily influence time-to-market, cost of capital and location decisions for first commercial plants.

Fusion sits at the intersection of energy security, decarbonisation and industrial policy, making it a natural candidate for blended public–private capital stacks.

For infra and energy LPs, track whether the EU converges on milestone-based, state-backed models akin to early US fusion programmes – likely a precondition for project-finance-style capital.

Distinguish between equity in fusion startups and eventual exposure to plants and equipment; technology, regulatory and financing risks differ sharply.

Map second-order exposure: superconducting magnets, specialised materials, tritium and lithium supply chains that will be needed if fusion scales.

Global private investment in space tech reached $12.4B in 2025, surpassing the 2021 peak and fully recovering from the 2022 downturn.

The U.S. led with $7.3B, driven by defence-linked satellite systems (including the Pentagon’s “Golden Dome”) and launch infrastructure.

A broader “space economy” measure – satellites, launch, ground infrastructure, data and defence resilience – captured $55.3B across 431 companies in 2025.

Confirms space as a structurally growing capital pool, not a post-SPAC fad: the sector is now outperforming broader VC on recovery metrics.

Most new dollars are going into communications, defence and infrastructure, not tourism or purely speculative plays, shifting risk and policy profiles.

A potential SpaceX IPO is widely seen as a validation and liquidity event that could re-rate late-stage space valuations and open IPO paths for followers.

Break out “space” exposure into launch, LEO broadband, EO/data, defence constellations and ground infrastructure – they have very different cycle and policy sensitivities.

Ask managers which data sources and definitions they use (Seraphim vs Space Capital vs in-house) before relying on headline funding totals.

Treat space as a distinct risk factor in portfolio construction; tens of billions per year now make “unseen” concentration unlikely.