Crowding In, Reversed: The Scaleup Europe Fund's Second Investment Went Into a Delaware Company

The Diagnosis Was Correct

Thanks to the Draghi Report, nobody in Brussels can pretend any more that Europe does not have a competitiveness problem. The report's call for Europe to close its deep-tech financing gap is now cited by the European Commission itself as the direct motivation for its newest financial instrument: the European Innovation Council describes the Scaleup Europe Fund as "a concrete delivery of the European Commission's Competitiveness Compass and a direct response to the Draghi Report's call for Europe to close its deep-tech financing gap."¹

The underlying diagnosis is not controversial. Europe has a major issue making funding available for startups. Compared with China and especially the United States, Europe is consistently behind on capital, and a large share of the European companies that do get funded are funded by American investors. The problem is not evenly distributed across the funding lifecycle, though, and this is the part that matters. Europe is passable at seed. It is tolerable at Series A. Where it collapses is at the top end: the Series E and Series F rounds, the ones that run to hundreds of millions or into the billions. That is the range European investors simply cannot serve at scale.

The EIC's own materials concede the point in almost the same language, noting that there is "no fund of comparable size that provides direct equity investments in European strategic technology companies at the growth and scaleup stages, forcing many to seek funding outside of Europe."¹ The Commission's legal text is blunter still, describing the fund as "responding to the urgent need for Europe to boost investments in scaleups and contribute to closing the late-stage funding gap in Europe."²

What the Gap Costs: Northvolt and Lilium

The consequences of that gap are not theoretical. Europe has watched two of its most prominent technology champions fail inside a short window. Lilium, the German electric aircraft company building small planes for vertical take-off and landing, and Northvolt in Sweden, which collapsed dramatically for want of funding.

There is a reflex that follows any high-profile failure, and it deserves to be named. As soon as a company fails, everybody looks like an idiot. People come out of the woodwork to explain that the founders were incompetent, that the technology never worked, that the whole thing was obvious in hindsight. Some of that criticism is fair in individual cases. But the simple truth is that with enough money, most of these companies would have done considerably better. Capital does not excuse mismanagement, and it does not turn a bad company into a good one. It does determine whether a capital-intensive hardware company gets to survive its own learning curve, which is exactly what battery manufacturing and aviation certification require.

Both failures hurt Europe in areas it can least afford to lose. Northvolt in particular struck at batteries and recycling, which is precisely the kind of strategic industrial capability the continent keeps saying it wants. These were also exactly the sort of rounds the Scaleup Europe Fund now exists to support.

The Playbook: Catalytic Public Equity

The Commission's answer was taken from a playbook the European Innovation Council had already been running. The EIC takes government money and deploys it into private companies as an equity investment, with the explicit goal of unlocking private capital alongside it. The public euro goes in first; four or five private euros follow; the public body then reports that it acted as a catalyst, mobilising far more capital than it committed.

The Scaleup Europe Fund is that mechanism, moved up a weight class. The idea was to apply the same catalytic logic to the very large late-stage rounds, the hundreds of millions the EIC's existing instruments cannot reach. The scale difference is explicit in the EIC's own comparison: its existing STEP call is capped at EUR 30 million, while the Scaleup Europe Fund targets investments "in the range of EUR 100 million and above."¹

The Structure and the Money

The fund sits under the existing EIC Fund umbrella but is run by a fully privately owned, market-based manager. Following a competitive public call that closed on 3 February 2026, the EIC Fund Board selected EQT as fund manager, a decision confirmed on 18 May 2026 and announced at the EIC Summit on 3 June 2026.^1,3

The headline numbers are worth stating precisely, because the round figures that circulate are not quite what the documents say. The European Commission has committed EUR 1 billion. That billion is itself staged: EUR 600 million from the 2026 Work Programme and a further EUR 400 million "subject to the 2027 EIC Work programme."² The billion is intended to attract EUR 4 billion from other investors, "to reach a capitalisation of approximately EUR 5 billion."²

The five billion is a target, not a balance. The public commitment is therefore roughly 20 per cent of a fund that does not yet exist at full size, and 60 per cent of that public commitment is the only part firmly appropriated. The founding investors alongside the Commission are Novo Holdings, EIFO, CriteriaCaixa, Santander/Mouro Capital, the Fondazione Compagnia di San Paolo/Intesa Sanpaolo/Fondazione Cariplo group, ABP represented by APG, Wallenberg Investments and Allianz.¹

On governance, the Commission "will participate on equal terms with the other investors and have a corresponding representation in the Fund's governance structure."¹ That sentence is easy to over-read. The same page removes the ambiguity: EQT operates as "an independent, market-based investment manager", and "the European Commission and other investors will participate in governance structures as investors but will not direct individual investment decisions."¹ The Commission is a limited partner with a seat at the table and no hand on the trigger. It is not entirely without recourse, though, and this is the detail to hold on to. The Work Programme reserves one specific power for the Commission: it "will reserve the right to refrain from any specific investment where the relevant proposed investment does not comply with applicable Horizon Europe requirements."² That is not a commercial veto and was never meant to be. It is a compliance veto, and it exists for precisely the case where a proposed investment fails the eligibility rules. Whether that separation is prudent or naive is precisely what the fund's first fortnight tests.

Eight Days, Two Investments

The fund became legally operational on 4 August 2026. On 5 August, one day later, the first investment was announced: ICEYE, the Finnish satellite operator, in a very large late-stage round. On 12 August, eight days after the fund could legally act, the second was announced: Lovable.

Hold that timeline in mind, because it is the single most revealing fact in this entire story, and we will come back to it.

Investment One: The Fund Working Exactly As Designed

The ICEYE investment is what the Scaleup Europe Fund was built for, and it is worth saying so plainly before criticising anything else. Run it against the mandate line by line. Is it a late-stage round of the size Europe cannot otherwise serve? Yes, a Series F reported at around one billion euros, which is a magnitude Europe rarely sees. Is the company established in an EU Member State? Yes, Finland. Is it a strategic technology? Satellite constellations are about as strategic as the category gets, and Europe is lagging badly in exactly that domain. Is it deep tech in any serious sense? Yes: genuinely differentiated technology, a real moat, heavily capital-intensive, hard to replicate.

Everything checks out. This is a bold investment that a European taxpayer can look at and understand. If the fund made ten more of these, the argument for its existence would make itself.

Investment Two: What Lovable Actually Is

Lovable is an AI coding startup, a vibe coding platform. You describe what you want and it builds it: a website, an app, a tool. It orchestrates models from various providers behind its own harness and interface, and it has extended into full-stack work, so you can build the front end, the back end, the integrations, and then push, launch and host the result.

It is, by any commercial measure, an outstanding company. The trajectory is genuinely remarkable: a USD 7.5 million pre-seed in October 2024 led by Hummingbird and byFounders, a USD 200 million Series A at a USD 1.8 billion valuation led by Accel in July 2025, a USD 330 million Series B at USD 6.6 billion in December 2025 led by CapitalG and Menlo Ventures with Databricks, Salesforce Ventures, Khosla, DST Global and Atlassian participating, and then the USD 400 million Series C at USD 13.3 billion in August 2026.⁴ That is more than USD 930 million raised across four rounds in under two years. Revenue followed the same curve, from roughly USD 200 million ARR in late 2025 to roughly USD 600 million by August 2026, and the company was adding revenue at a rate that made headlines even at 146 employees in March 2026.⁵

None of this is a criticism of the company. Great execution, great team, a real niche captured at speed. The objection that follows is not about whether Lovable is a good business. It is about whether a European public instrument with a strategic mandate had any business being in that round.

Why It Is Not Strategic

Start with the competitive position. Every major model provider is moving into this exact layer. OpenAI has Codex. Anthropic has Claude and its agentic coding surface. Google built its own. These are the companies that own the models Lovable depends on, and they are now shipping the tools that sit on top of them. The cake is getting smaller, and the people slicing it also own the oven.

That structural point matters more than any individual quarter of growth. A vibe coding platform that does not produce its own model is permanently paying a margin to whoever does. The more the model providers compete directly in the application layer, the harder that position becomes to defend. Lovable will very likely be fine, and will be around for a long time. But "will be fine" is not the same as "strategic asset for a continent."

Now ask what actually is strategic for Europe right now. Energy costs. AI infrastructure. Data centres. Foundation models and the model providers themselves. Hosting. Those are the chokepoints where dependency is real and expensive. Agentic coding is not one of them, because the models are not developed by Lovable. They are developed by someone else. From a sovereignty standpoint, what difference does it make whether a European developer reaches OpenAI and Anthropic through Lovable or through some other client? The dependency is identical either way. The strategic exposure sits at the model and infrastructure layer, and the Scaleup Europe Fund did not buy anything there.

Put more directly: this is an extremely well-executed wrapper around AI. That is a compliment about the execution and a statement of fact about the architecture. It is not deep tech in the sense the fund's own documents use the term, which runs to scientific, engineering-heavy, IP-heavy technology. Lovable is not making moves into AI infrastructure, hardware, data centres or foundation models. It has partnerships, of course, but it is not entering the hyperscaler game. With its execution and growth it could conceivably build toward that, but it is not on the roadmap today.

The Delaware Flip

Then comes the part that genuinely changes the character of the transaction. Lovable is described everywhere as a Swedish company. Its parent entity, Lovable Labs Inc., is incorporated in Delaware. The Swedish operating entity, Lovable Labs Sweden AB, sits underneath it. Stockholm holds the headquarters, the workforce and the R&D. Delaware holds the company.

This is not a scandal on Lovable's part. It is a completely standard manoeuvre, and the reason it is standard is a European failure, not an American trick. Suppose you have a company in Sweden or in Germany and you want US investment. Many US late-stage investors will not write nine-figure cheques into a continental European entity, and the reason is procedural absurdity. In parts of the EU, an investment round means signing twenty copies of the agreement, and getting every signatory and every investor into a room with a notary who reads every single word of a fifty-page document aloud to everyone present before anyone signs. In Germany you genuinely have to do this. It is a ridiculous amount of bureaucracy, and there are several EU countries where some version of it applies.

So investors say no. Instead: create a holding company, put the existing company underneath it as a subsidiary, and we will invest into the holding company in the United States. That is the Delaware Flip. Not everybody does it. A great many companies do.

Here is why it matters for this particular investment. After the flip, the investors do not sit on the Swedish cap table. They sit on the American one. Which means the Scaleup Europe Fund did not invest in a Swedish company. It invested in a US company. In a USD 400 million round, whatever the fund contributed, whether tens of millions or more, that money went into a Delaware entity.

Testing the Possible Justifications

There are only a few reasons a strategic European fund would knowingly do this, and each falls apart under pressure.

Reason one: bring the holding company back to Europe. A reverse flip, moving the Delaware parent back to Sweden. This will not happen. There is far too much American money on that cap table. It would be extremely tax inefficient, very expensive and, from the shareholders' perspective, straightforwardly stupid. The US holding structure is going to stay exactly as it is.

Reason two: employment. Perhaps the Commission wanted to protect European jobs. The numbers do not support it. Lovable had 146 employees in March 2026,⁵ passed 200 later that year according to co-founder Anton Osika,⁶ and is targeting around 450.⁷ So we are in the hundreds. Not the thousands, not the tens of thousands. Lovable is not a significant employer, and no plausible reading of European employment policy justifies a strategic fund's cheque on those grounds.

Reason three: blocking an acquisition. This is the only one with any substance. If an American acquirer comes for Lovable, perhaps the fund wants standing to object. The difficulty is arithmetic. The fund was one investor among many in that round, and its individual cheque size has not been disclosed. Suppose it took half the round or less. Of the fund's own capital, only about 20 per cent is public money, since the remainder comes from the private founding investors. The genuinely public share of a USD 400 million round is therefore a small fraction of a fraction. How much blocking power does that buy? Almost certainly none, particularly since the Commission has explicitly agreed not to direct individual investment decisions.¹ And there is a second obstacle that has nothing to do with cheque size: any such right has to be negotiated against co-investors who have every financial reason to refuse it, because a veto that blocks an American acquirer also caps the price everyone else on the cap table can be paid. Menlo, Tencent, Balderton and the rest did not write into Lovable in order to have their best exit foreclosed by somebody else's policy objective. Whether the fund negotiated any exit rights is unknowable from outside, and that opacity is itself worth examining. The Work Programme says the specific terms of each investment "will be considered and negotiated on a case-by-case basis in accordance with the Scaleup Europe Fund Investment Guidelines to be adopted by the EIC Fund prior to the commencement of operations," and promises that "access to the Scaleup Europe Fund Investment Guidelines will be made available on the EIC website."² A rulebook was therefore supposed to exist before 4 August, and outsiders still cannot read it against the Lovable deal.

The safeguard regime is worth following, because tracing it produces the sharpest confirmation of the whole argument. The fund's own section does not spell its safeguards out. It defers: "economic security safeguards, in line with the provisions in the Introduction part of this Work Programme, section 'Economic Security', will be addressed through the investment guidelines and the terms of the investments."² So the safeguards live in the unpublished Investment Guidelines, and the only published articulation of what they are sits in the Introduction, written for a different pathway. There, the trigger is a Commission Award Decision, and safeguards attach where that decision identifies the need.² The Scaleup Europe Fund has no Award Decision at all: applications "will be submitted directly to the Scaleup Europe Fund and will be assessed and decided by the independent investment adviser and portfolio manager."² The published mechanism has no place to attach.

Now read the scope limit, which is the part that matters. Safeguard measures apply "only in the case of projects (i) selected under the STEP Scale up call or falling within the four priority technology areas and (ii) involving specific technologies listed in the second column of the Annex to Commission Recommendation C(2023)6689." Artificial intelligence is one of those four areas, so a reader might assume Lovable qualifies. The next sentence forecloses it: "the safeguard measures will only concern projects whose core purpose is to develop one of these specific technologies and not projects where these technologies are a tool or method that will be applied for a specific application, but without substantial development of the technology itself."²

Read that against Lovable and it is almost uncomfortable. The Commission drew a line between developing a strategic technology and applying one, and put its protections only on the first side of it. Lovable is on the second side. It does not develop the models; it applies them for a specific application. By the Work Programme's own test, this is not the kind of company the economic security architecture was built to protect. The fund invested in it anyway. Unknown is the honest answer on exit rights, and unknown is not a strategy, but the safeguard text is not silent. It quietly says this company was never the point.

The Fund's Own Words, Read Against Its Own Deal

The most damaging material here is not external criticism. It is the fund's own published mission, read line by line against what it did eight days into its existence.

On foreign dependency, the EIC writes that the fund "will also help address the increasing dominance of foreign investors in large funding rounds, thereby reducing risks such as strategic dependency, company relocation, foreign control, and talent outflow."¹ Take those four risks individually. Company relocation: the company has already relocated its ownership. Foreign control: it is a US entity, and the round was co-led and heavily populated by US investors. Strategic dependency: the platform depends entirely on American foundation models. Talent outflow: possibly the one item still live, and only if the fund holds rights it probably does not.

On who may invest, the fund is open to institutional investors "that share the objectives of SEF, to retain strategic assets in Europe and create a sustainable deep-tech ecosystem."¹ Retaining a strategic asset in Europe is difficult when the asset's parent is in Delaware, and the deep-tech characterisation does not survive contact with the product.

On eligibility, the EIC page says the fund is open to companies "located in or intending to locate to any EU Member State or countries associated with Pillar III of Horizon Europe."¹ The phrase "or intending to locate to" is doing an enormous amount of work, and is presumably the hook on which this deal hangs.

Then there is the Commission Implementing Decision, C(2026)2051 of 26 March 2026, which is the binding text rather than the marketing page. It states that the fund "will ensure that supported companies retain the majority of their value creation, including intellectual property, within the EU or Horizon Europe Associated Countries, and give preference to European exits to investments, thereby contributing to technology leadership, economic growth and job creation in Europe."² Preference to European exits means that if someone acquires you, it ought to be a European acquirer, not an American or Chinese one. Job creation, as established, is not the story here.

And most pointedly, on who may receive the EU contribution: "investments will be limited in all cases to single companies qualifying as SMEs or small mid-caps (up to 499 employees), established in an EU Member State or in a country associated to the EIC investment component of Horizon Europe."²

The employee count is satisfied. The establishment requirement is the problem. The United States is not a country associated to the EIC investment component of Horizon Europe. It is emphatically not. The associated third countries are a short list: Norway, Switzerland, and others such as Ukraine and Turkey. The US is not on it, and was never going to be.

So on a plain reading, European public money went into a company established in Delaware, in apparent contradiction of the fund's own eligibility rule. And this is where the reserved power from earlier matters, because the very same paragraph that sets the eligibility rule closes by reserving the Commission's right to refrain from any investment that does not comply with Horizon Europe requirements.² The one lever the Commission kept for itself was built for exactly this situation. It was not pulled.

There is a caveat worth stating fairly. The Work Programme does contemplate holding structures, though not in the Scaleup Europe Fund's own section. The eligibility rules for the EIC STEP Scale Up call, a separate and far smaller instrument capped at EUR 30 million, provide that an eligible company "may have a holding entity for the purposes of the investment, and this holding company must also be established in a Member State or an Associated Country."² If anything that cuts the other way. The one place the EIC framework addresses holding companies head on, it anticipates them and still requires them to be European. It does not rescue a Delaware parent.

Nobody Is Going to Enforce This

And of course nobody will do anything about it. Nobody is going to say, wait a second, you broke your own rules. Who has the money to sue the European Union over an eligibility clause in a Work Programme, and what would they even be suing for? It would be ridiculous. In fairness, one can also take the permissive view: let them cook. If the Commission wants to put the money there, this is in some diffuse sense what European voters signed up for. But it remains a little bit ridiculous that a fund created to keep strategic value inside Europe made its second investment into a US entity.

What a Strategic Second Investment Would Have Looked Like

The counterfactual is not hard to construct, because the fund had already demonstrated it the week before. Another satellite operator on the ICEYE model would have been defensible instantly. A company running data centres in Europe, or a startup building them, would have made obvious sense given that infrastructure is the actual chokepoint. Energy. Quantum computing. Any of these would have justified the cheque and the fanfare.

For a vibe coding platform, going out of your way to invest in a US entity, where no strategic rationale is visible, the risk profile is inverted. What is the best case? And what is the worst case if it is acquired by an American company, which is honestly the most likely outcome? Would Europe actually miss out because it lost Lovable? Probably not. Lovable will most likely do very well in the United States, quite possibly better than it would have in the EU. The strategic importance is not there in either direction.

A Prediction About the Coding Layer

It is worth stepping back to the industry dynamics, because they explain why this asset is likely to leave anyway.

The hyperscalers, Microsoft, Meta, Google, are all trying to win the next category, and they are all moving toward where the money will be. That means energy, data centres, and either owning models or partnering closely with those who do. Different players are ahead in different places. Data centres are the clearest battleground, with Google, Microsoft and Meta all moving hard, and Amazon and Microsoft already deep into it, while Meta scrambles to follow because it recognises both the opportunity and the necessity of winning the infrastructure game.

The same consolidation is happening one layer up, in the coding and agentic tools. Google shipped its own. The model providers keep launching into the space. And the largest signal of all: Cursor, the AI-native development environment, essentially a reimagined VS Code aimed at professional developers rather than at Lovable's more general audience, was acquired by SpaceX in an all-stock deal valued at about USD 60 billion, agreed in June 2026 and closed on 14 August 2026.⁸ SpaceX is itself moving toward data centres, and evidently concluded that the coding surface matters to winning the agent game.⁹ Cursor had roughly USD 3 billion in ARR and a prior valuation near USD 29.3 billion, which is the scale of revaluation this layer is now seeing.

The pattern is consistent: own the infrastructure, own the tokens, and own the tools people use to spend them. We give you the compute and inference, and we give you the interface that consumes it.

Meta is the conspicuous laggard. It is only now moving seriously into data centres, and it needs a strong coding product. Meta's historical playbook is to buy rather than build, and when its own version underperforms, it acquires the leader instead. So the prediction is straightforward: an acquisition of Lovable, or of a competitor like Replit, valued at around USD 9 billion after its March 2026 round, would make a great deal of sense for Meta. It would be expensive, and Meta is currently spending heavily on data centres, so it may not want the cheque right now. But as an exit path it is entirely plausible, and the shrinking economics of a model-less coding platform make an exit more attractive over time, not less.

If that happens, the Scaleup Europe Fund will have used European public money to help capitalise a company shortly before an American acquirer took it, having secured no demonstrable ability to prevent that outcome.

The Timeline Is the Real Story

Return to those eight days. The fund became operational on 4 August. Its first investment was announced on 5 August, its second on 12 August. Both were very large rounds, roughly one billion euros for the first and USD 400 million for the second.

You do not source, diligence, negotiate and paper rounds of that size in a week. It is not possible. Which means the deal flow existed before the fund was operational. These were rounds already being assembled by the asset manager, into which capital was then placed.

That reframes the whole exercise. How much genuine due diligence did the Scaleup Europe Fund perform on the policy dimension, as opposed to the commercial one? Very plausibly not much. The manager had the pipeline and was looking for capital to deploy into it. The fund had capital and a mandate to deploy it quickly. The policy screen, the part that asks whether this specific company advances European strategic autonomy, appears to have been the thinnest layer in the stack.

It also suggests motive. This looks like an overreaction to Northvolt and Lilium. Those failures hurt, and they hurt in areas that matter enormously, batteries and recycling and advanced manufacturing. The institutional response to that pain has been to invest quickly in anything that can be described as strategic, rather than slowly in the things that actually are.

Who Crowded In Whom

The stated theory of the Scaleup Europe Fund is catalytic: public capital goes in first and crowds in private capital behind the European public interest. Here is what appears to have happened instead.

Lovable's Series C was reportedly in the market at USD 300 million at a USD 13.2 billion valuation before it closed at USD 400 million at USD 13.3 billion.^10,7 The round was upsized. Menlo, Tencent, Balderton, Carmignac, Kaszek and Regent all wanted in. By any reading, this was not a company that could not raise money in Europe and was therefore forced abroad, which is the precise scenario the fund exists to address. Lovable did not need the Scaleup Europe Fund's money at all.

So the private investors did not follow the EU into its mission. The EU followed the private investors into their deal. The fund was not the catalyst; it was the last cheque into an already oversubscribed round in a hot category, placed into a Delaware entity, eight days after it was allowed to write cheques at all.

The EIC's stated purpose is to crowd private capital into a public agenda. What happened here is the inverse: private funds had the deal flow, and they crowded government money into theirs. It does not look like private investors are following the EU's mission. It looks like the EU is paying into the private funds so they can follow their own.

Let us see whether that changes. The fund is two weeks old and one bad fit does not condemn an instrument that Europe genuinely needs. The ICEYE investment proves the thing can work as intended. But the second decision revealed how the machine behaves under commercial pressure, and the answer is that the policy mandate yields. If the next ten investments look like the first one, this will read as a stumble. If they look like the second, then Europe has built a EUR 5 billion vehicle for subsidising other people's deal flow, and has done it with the language of sovereignty printed on the outside.

Works Cited

1. Scaleup Europe Fund - European Innovation Council, accessed on August 19, 2026
2. Commission Implementing Decision C(2026)2051 of 26 March 2026 amending Implementing Decision C(2025)7410 as regards the introduction of the contribution of the EIC to Scaleup Europe Fund, with the annexed EIC Work Programme 2026 (Section IX.1 Scaleup Europe Fund; Introduction, section Economic Security; EIC STEP Scale Up call), EUR-Lex, accessed on August 19, 2026
3. EQT selected to lead the Scaleup Europe Fund - EQT Group, accessed on August 19, 2026
4. Vibe-Coding Startup Lovable Hits $13 Billion Valuation - The Wall Street Journal, accessed on August 19, 2026
5. Lovable says it added $100M in revenue last month alone, with just 146 employees - TechCrunch, accessed on August 19, 2026
6. Anton Osika: "The Lovable team just passed 200 people" - LinkedIn, accessed on August 19, 2026
7. Lovable headcount target and Series C terms - The Wall Street Journal, accessed on August 19, 2026
8. SpaceX to buy Anysphere (Cursor) for $60 billion - Reuters, accessed on August 19, 2026
9. SpaceX model training partnership - Cursor, accessed on August 19, 2026
10. Lovable in talks to raise $300m at $13.2bn valuation - Sifted, accessed on August 19, 2026