SuperSeed's March 2026 briefing collects five debates from inside the firm, from whether AI start-ups are really services businesses to Europe's AI dependency, two-tier VC pricing, services-as-software and AI-attributed job cuts, each with a hot take and reading list.
- Fast-scaling AI start-ups average only around 25% gross margins, against the 80-90% SaaS margins that VC return maths is built on.
- Europe's annual AI investment is only 10-15% of US levels; SuperSeed's view is that Europe cannot win the foundation model race but can win in application and vertical AI in regulated industries.
- For seed funds the briefing argues for pricing discipline: avoid the AI news cycle and AI wrappers, and lean into applications and vertical AI in Europe.
- The services-as-software thesis frames AI's market as the roughly $6 trillion global professional services market rather than the software market.
- SuperSeed reads Block's 4,000 job cuts, attributed to AI, as more likely about pleasing investors and returning to pre-bloat headcount.
In the UK Spring is starting to spring. Can you feel it? A rebirth.
With that in mind, here all some key conversations from behind the scenes at SuperSeed towers.
The Slack chats, IC musings, watercooler moments that have been grinding gears…
…the trends, the themes, the challenges, the opportunities – connecting the dots for founders, partners, LPs and GPs we work with, across the worlds of physical AI and UK & European sovereignty.
1. Are AI Startups “Services Businesses” & Should We Price Them Accordingly?
Context: Fast-scaling AI “Supernova” Startups are averaging only 25% gross margins, fundamentally different from the 80-90% SaaS model that VC return maths is built on. While revisiting the “SaaSpocalypse” and “SaaS is dead” conversations we have quite regularly (Good SaaS is not dead btw).
The hot take: Are VCs underwriting AI companies at software multiples while they have services-business economics? (Cursor was paying roughly $650 million annually to Anthropic while generating approximately $500 million in revenue, a negative 30% gross margin. According to third-party analysts)
To Read:
The AI Churn Wave – Chart Mogul
Upgrading SaaS models for AI – from McKinsey
“This isn’t our 1st SaaSpocalypse” – Marc Benioff
2. Is Europe’s AI Dependency Worse Than Its Defence Dependency?
Context: Europe’s total annual AI investment is currently only 10-15% of US levels, and the gap is widening, not narrowing. The US has produced 40 AI foundation models, China 15, and all of Europe combined, err, 3.
Does sovereign AI matter or can we just bolt on Chinese OS and so what?
92% of Western data is stored in the United States, and EU attracts just 7% of global AI investment versus 40% for the US.
The hot take: This dependency is actually a massive opportunity for investors. We can’t win the foundation model race, but application layer and vertical AI we can. Mainly across regulated industries – healthcare, financial services, manufacturing, defence etc.
This is where European startups have structural advantages, domain expertise, regulatory fluency, and access to the actual enterprises that need AI deployed. Amplified by the “**** Trump Effect”.
To read:
Decoupling from Trump’s America – FT
Tech sovereignty comes at a price – FT
Europe’s declaration of independence? – Atlantic Council (Report)
Who build the best AI models? – Index (not that one)
3. Is A Two-Tier VC Market Creating a Once-in-a-Decade Opportunity for Disciplined Investors?
Context: AI startups are commanding significantly higher valuations and round sizes, and the hyperfocus on AI has had widespread impacts on fundraising for *real AI businesses* plus other sectors. Where is Alpha hiding *really*?
The hot take: It’s all about pricing discipline for seed funds. Stop chasing the ‘news-cycle’ AI hype-cycle. There are AI startups and AI startups. Lean into applications and vertical in Europe. Let everyone else duke it out for AI wrappers because that day has passed.
To read:
How AI Startups juice up valuations – WSJ
A 2-tier VC market – Pitchbook
VC outlook for 2026: 5 key trends – Harvard Law
4. Is The ‘Services as Software’ Thesis Real? Does It Invert Everything We Think We Know?
Context: The “services as software” thesis i.e. AI that delivers outcomes rather than selling seats (mainly in professional services) *could* be a solid frame for B2B investing. OpenClaw is creating a wave. Will enterprise buy or build? Will SaaS become just a system of record?
The hot take: AI can already do the work of a full-service consulting team, the TAM isn’t the software market it’s the professional services market (~$6 trillion globally).
To read:
How to run OpenClaw safely – MS
Value & Agentic AI – McKinsey
Consultancies must become software companies. – FT
Services as software: a review – Workpath
5. Will AI Take Your Job? Or Will Somebody Using AI?
(CEOs *Will* Use It As An Excuse, For Now)
Context: watching Jack Dorsey from Block cut 4,000 jobs and using AI as the excuse made us ponder once again what kind of impact AI will have on people, especially those in white-collar roles. The Citrini research paper, fresh in our minds.
The hot take: The reality with Block is more likely him pleasing investors (stock jumped 20% BUT still down 72% from 2022 high), cutting back to ‘pre-bloat’ employee numbers. And they’ll more than likely follow Klarna, who cut, and are now rehiring realising that AI isn’t the utopia they hoped for.)
To read:
The Consequences of Abundant Intelligence – Citrini
Blocks Job Cuts – Guardian
It’s AI’s “potential” – not reality – HBR
Frequently Asked Questions
Should AI start-ups be valued like software companies?
The briefing questions it. Fast-scaling AI start-ups average about 25% gross margins, closer to services economics; it cites third-party estimates that Cursor paid roughly $650 million a year to Anthropic while generating about $500 million in revenue.
Where can European AI start-ups compete, according to SuperSeed?
In the application layer and vertical AI, mainly across regulated industries such as healthcare, financial services, manufacturing and defence. European start-ups have domain expertise, regulatory fluency and access to the enterprises that need AI deployed.
What is the services-as-software thesis?
It describes AI that delivers outcomes rather than selling seats, mainly in professional services. On that view the addressable market is the professional services market, about $6 trillion globally, rather than the software market.