Upside is a weekly podcast designed to look behind the headlines that will affect European venture, startups and investing.

Below are the notes from this week’s episode. Episode links above to tune in and stream wherever you pod.

We recorded this on Friday 23 October, five days before the Budget. Mads Jensen and I were joined by Karim Palant, Director of External Affairs at UK Private Capital (formerly the BVCA), the trade body for Britain’s private equity and venture capital industry. Karim runs its public affairs, communications and research. Before that he ran public policy at LinkedIn and Facebook, and before that he was Ed Balls’ head of policy.

So yes, we had someone who actually knows how the Treasury sausage gets made. We covered the Budget, where money for British tech should come from, the financing wobbles in the AI build-out, and a rough week in cyber.

The warm-up: narrative, gas and giveaways

Mads has a rule of thumb for earnings. If you have free cash flow, you lead with it. If not, EBITDA, then margin, then revenue. And if you have none of those, you do narrative. Tesla’s Q3 was $28 billion of revenue, flat on a year ago, with operating profit down by more than 50% to $718 million. In Mads’ words it was “narrative heavy, results light”.

On Jackdaw, Karim’s read is that the decision is mostly about political handling. The manifesto promised no new exploration, but Jackdaw and Rosebank were approved by the previous government and then blocked by legal challenge. That gives Labour its hook. The energy-wealth-fund talk is there to head off the old “Thatcher squandered North Sea oil” argument from the left.

Then there’s Anthropic putting 5% of its shares into a public benefit fund. At the $2 trillion valuation being talked about, Mads puts that at around $100 billion. He called the team “almost a bit of a cult” (I think that’s too strong), but he also called it a powerful signal that nobody else has matched. Karim had a better takeaway: it proves the public affairs guy can be a co-founder.

A boring Budget, on purpose

The run-in has been quieter than in 2024 or 2025, Karim says, and that’s by design. Rumours were kept tight and the Budget was talked down as steady-as-she-goes. The previous chancellor left £24 billion of headroom, so something like the Straits of Hormuz can go wrong without forcing dramatic tax moves.

The industry’s line has been “do no harm”. No CGT rise and no fiddling with carried interest again. A proposal to equalise CGT with income tax has already been ruled out, and on-record briefing has ruled out CGT rises more broadly. Karim also points to something I hadn’t clocked. Healey keeps saying profit out loud, and previous chancellors didn’t routinely do that.

There are two carrots on the list. The first is a scale-up reinvestment relief for founders who recycle their exits. The second is more financial-transaction capacity for the British Business Bank, so it can cornerstone larger, later-stage funds and fill the gap the European Investment Fund left behind. If that doesn’t come through, Karim warns, the BBB will “rein in their horns” just as UK venture and AI have real momentum.

I was expecting less carrot, honestly.

Mads: fix the deficit, the rest is tinkering

Mads didn’t buy the idea of a silver bullet. In his view, the best thing Healey could do for founders is get the public finances under control. Spending has gone from 39% to 45% of GDP since Brexit. The civil service went from under 400,000 in 2016 to more than half a million today. Debt has passed £3 trillion. Heavy government borrowing pushes up rates, and higher rates make borrowing dearer for businesses and households.

Karim agreed with the broad sweep. He quoted George Osborne calling it “an interesting choice” to go boring at the point of maximum power. But he thinks Healey and Andy Burnham are laying the ground on welfare by talking about youth unemployment and the incentives that keep people out of work, which is the argument Starmer never made. Whether they win it is another matter.

Where British tech money should come from

This was the bit I really wanted Karim for.

About 85 to 87% of the money in UK private capital funds is international in a typical year. For VC it’s only about 60%, which surprised me. The domestic 40% is mostly the British Business Bank plus wealthy individuals: GP commits, family offices and ex-founders.

Pensions are the gap. Overseas pensions put £13 billion into UK-managed funds last year, which is 16 times the total from every UK pension. Defined benefit schemes do some of this. DC, where almost all of us now save, does very little. The Mansion House Compact has crept from 0.36% to 0.6% to 0.85%. Karim also called out the industry’s “triumphalist” progress claims, because most of that money has gone into property under the separate Mansion House Accord.

It’s a security question as much as a quantum question. In one year a couple of years ago, more Swedish pension money went into UK venture and PE than UK pension money. If the EU changes the rules, we’re stuffed.

Both Karim and Mads were clear that mandation is the wrong answer and incentives are the right one. Mads points to the plumbing. Final-salary schemes have to value liabilities off bond yields, so they get no credit for taking risk. DC schemes chase fees, which means index funds, which means large-cap US stocks. “We’ve designed the system to do exactly what it’s doing.”

Karim’s NOVA proposal is effectively a pre-diligenced list of UK funds, put together for pension trustees who keep claiming there’s nothing investable. Then he told the story of shopping for UK Private Capital’s own staff pension. The advisers pitched low fees and a nice app. When he asked about returns, they said they’d go and look.

AI build-out: wall or hiccup?

Microsoft walked away from its anchor-tenant deal with Nscale because grid power won’t arrive until the 2030s. Meanwhile Nscale is in talks with Anthropic about $45 billion of compute from West Virginia. At least 47 large US data-centre projects have been delayed, withdrawn or blocked. Default insurance on AI debt hit $9 billion traded in September, a big jump from July. And Meta is still guiding to $135-145 billion of capex.

Mads is firm that this isn’t a wall. There are no dark GPUs, capacity is sold out, and it costs more to rent an old H100 now than it did six months ago. The neoclouds are getting rickety, and some may fold into the hyperscalers. OpenAI’s pre-IPO round has been marked down from $1.4 trillion. For Mads, the real question is velocity. I think a lot of that $1.2 trillion is PR.

PHP, AI and the cyber week

PHP is the P in LAMP (Linux, Apache, MySQL, PHP) and still runs about 70% of websites. Mads explained that the flaw has been known for around 20 years, but stringing the exploit together was too hard. Once a human pointed AI at the right weakness, it brute-forced the chain in under a week. Before AI, that would have taken the best people months.

Alongside PHP, LiteLLM, Railway, Zeabur, OpenCode, Microsoft 365 and about a hundred companies, including John Deere and Cummins, were broken into by AI agents for around 50 bucks a pop. The attackers stole AI keys and used the free compute to attack again. Cybercrime is now almost self-funding.

Mads’ analogy is getting your driving licence at 18. You’re less safe, but you can do far more. Karim’s worry is the people making the calls. Few politicians he’s known have the expertise, so they lean heavily on NCSC and GCHQ. Paying 70 grand for a head of AI won’t fix that.

Predictions

• Me: Next year we’ll see major security breaches around personal agents, especially on the financial side.
• Me: Far less than the $1.2 trillion of projected AI capex will actually roll out.
• Mads Jensen: “I don’t think it’s a wall we will be keeping build out.” The question is velocity.
• Karim Palant: “Relatively optimistic there’ll be something” in the Budget on entrepreneurs exiting and reinvesting.
• Karim Palant: At the current rate, the Mansion House 5% target lands about ten years late.
• Karim Palant: On welfare, “if they don’t grip it, it will get them in the end.”

Deals of the Week

• Photon Bridge (Eindhoven), €10 million round – picked by Mads. Data centres move so much data that copper isn’t fast enough, so they need light, and light needs lasers. Photon Bridge has found a better way to make data-centre lasers. Mads called it a hat-trick for the Brainport cluster after EUCLYD. He traces that cluster back to the old Philips mafia that spun out ASML.
• Karim’s alternative pick: The final Budget has already gone to the Office for Budget Responsibility, so “whatever deal the British Business Bank has got” was settled earlier this week.

Notable Quotes

“it shows that the policy and the public affairs guy can be a co-founder, which as a policy and public affairs guy is something to aspire to.” – Karim Palant

“Like the value of the pension today versus yesterday is totally irrelevant. What are the returns? And they said, that’s a good question. We’ll go and have a look. We’ll see if we can find that.” – Karim Palant

“Now, what happens if the EU changes the rules and makes that impossible? We’re not in control of our destiny.” – Karim Palant

“So my definition of a wall is something that stops you abruptly and immediately.” – Mads Jensen

Why This Matters

UK VC funds still depend heavily on foreign LPs and one government bank. The Budget tells us whether the British Business Bank keeps scaling or reins in. The pensions fight tells us whether UK savers ever own a slice of UK upside, or whether US investors keep taking the returns when our companies leave. For GPs raising now, both matter more than any headline tax rate.

And on cyber, if a 20-year-old flaw falls in a week, every founder running old kit has a patching problem.

Boring Budgets are fine. Boring pensions are the expensive bit.