This week on Upside: Nvidia is so rich it's literally throwing money at shareholders, three megacaps queue up to IPO, and Americans discover they hate AI almost as much as politicians.

(00:47) Mads at the BBCA / UKPC. Same name, same problem, same answers nobody's implementing. Plus a quick spin through Ben Evans' latest, ASML's revenue upgrade, and Meta canning 8,000 engineers. Cool, cool, cool.

(02:30) Nvidia returns $80B to shareholders - roughly the GDP of Tunisia, give or take. Free cash flow of $49B a quarter. Stock down on earnings - maybe we've collectively lost our minds.

(07:00) The Great IPO Unfreeze. SpaceX at $2T (an AI company wearing a rocket suit wearing a telecoms business), OpenAI at $1T (the most consequential S-1 of the decade, Dan bets it ain't pretty), and Anthropic's "final-final" pre-IPO round, number four. Sure.

(13:24) Unitree IPOs in Shanghai at $67B. The dancing robot company actually shipping units. The Android of humanoids? Europe, please build the layer on top.

(20:00) Europe writes its own future. Mistral grabs Vienna's EMMIE for neural physics simulation, EQT wins the €5B Scale-Up Europe mandate (sorry Atomico), and HMRC picks Quantexa over Palantir for £175M. Buying British: hopefully not the B-player.

(27:30) American AI rebellion. AI now polls worse than ICE. Molotovs at Sam Altman's house. Europe, take note.

(33:30) Prediction: France will cave on ESOP rules by end of '26. Mads disagrees.

(36:30) Deal of the week: Isomorphic Labs, $2.1B Series B. Demis's side hustle is going great.

(38:50) Week ahead: Salesforce earnings (where art thou, Agentforce?), AI-stack reports, and the S&P 500 rule change that could let SpaceX/OpenAI/Anthropic into the index faster than feels reasonable.

Recorded live from a Montenegrin port at a European LP/GP gathering of 80 investors, Dan asked 5 European investors the same question:

If you had to put every penny of your fund and personal capital into one bet in Europe, where would it go?

(01:21) Jens Lapinski — German super angel

Jens would back the full AI stack, money transfer infrastructure, training environments, enterprise deployment. On sovereign AI, he argues the US-Europe relationship is one investment continuum rather than two competing blocs: chips, foundation models, hyperscalers, operating systems and the application layer are predominantly US-built, and pretending otherwise isn't productive. He acknowledges defence and security carve-outs make sense, but the broader play is co-investing with American capital, not against it. The one structural fix: unlock European pension fund and insurance capital flows into venture, and fix the public market exit path so companies don't default to US listings.

(07:24) Andra Bagdonaite — First Pick, Lithuania

First-check pre-seed investor for Baltic founders since 2022. Andra's bet: AI applied to traditional SMBs; logistics, manufacturing, businesses still running on legacy ERPs and Excel. Lithuania functions as a fast sandbox for founders to pilot with one or two customers before going global. She frames this as a long compounding opportunity rather than a rocket ship, the scale of legacy process digitisation across European SMBs is the prize.

(12:21) Ella Goldner — Zinc

Ella would concentrate on transformative deep tech climate. Her view: incremental solutions won't shift the trajectory, so a portfolio approach to moonshot bets, fusion and energy provision featured explicitly, is the rational play. She cites Kate Bingham's COVID Vaccine Taskforce as the model: urgency strips out bureaucracy and accelerates the research-to-commercialisation path. Insurance pricing is already starting to reflect climate risk, which she sees as an early commercial signal. Her one change for Europe: cultural rather than structural, building a problem-solving, business-building mindset from school age. She contrasts this with Israel and the US, where building and discussing ventures is embedded in everyday conversation.

(23:58) Mike Reiner — 432 Legacy, Netherlands

Mike's bet: the software and safety layer around physical AI. Roughly 11% of revenue is currently lost to safety-driven slowdowns and caging where robots and humans share environments. His firm has already backed a Swiss company using ultra-wideband to detect people through walls for multi-sensor safety platforms. The thesis comes from a proprietary value-chain analysis system: the same tool surfaced transformer and specialist hardware supply as the binding constraint on data centre buildout (not just energy), and identified firmware security (portfolio company Binary) as a durable cyber gap less likely to be absorbed by the foundation model providers. On founders, he looks for trauma-driven ambition paired with active self-development work, the combination he sees correlating with mission-driven, durable companies. His one change: more funds willing to take on tough problems with longer fund cycles, which he argues is also where the largest demand-supply gaps, and returns sit.

(34:39) Dave Haynes — FOV Ventures

Dave's bet: robotics and physical AI. He separates the two cleanly; robotics is the embodiment, physical AI (or world models) is the intelligence layer that understands real-world physics rather than text and images. Sizing: digital economy roughly $15T, physical economy in the hundreds of trillions, and software has only meaningfully disrupted the former. FOV maps the space across four layers: hardware and infrastructure (sensors, hyperspectral imaging, wireless power, portfolio company Willow), software and operating systems, the intelligence layer (world models, where Yann LeCun's Amy sits as a European entrant), and vertical applications. Europe's edge, in his view, is in vertical deployment; labour shortages, real industrial pull, and strong robotics talent from TUM, ETH and Imperial. He frames the shift as "Robotics 2.0": where Robotics 1.0 needed £5m of CapEx, founders can now build serious automation businesses on a £500k cheque, mirroring the Web 1.0 to 2.0 transition. He estimates the ChatGPT-equivalent moment for robotics is roughly five years out.

Hosts: Mads Jensen (moderator), Lomax Ward (Lisbon beach bum), Dr. Andrew Scott (serial mute-button offender) Missing: Dan Bowyer — scaling actual mountains. We told him to come back alive. Mostly because nobody else wants to moderate.

[02:30] — ARM vs Spotify: One the Market Loves, One It Doesn't Both posted good numbers. ARM got rewarded, Spotify got a -13% haircut and a 25% YTD drawdown despite record margins and 761m MAUs. Lomax calls ARM's AGI CPU pivot a "bet the farm." Mads drops the CPU:GPU ratio shift from 1:8 to potentially 1:1 as agentic AI explodes tool calls. Infrastructure compounds, consumer optimises, and the market knows which is which.

[10:38] — Anthropic: $200bn Google Deal, $45bn ARR, Just Give Them Europe Too $200bn to Google Cloud. ARR at $45bn (up from ~$9bn in January). Potential $50bn raise north of $1tn. Lomax declares Anthropic "so, so hot right now" — secondary-market appetite lives in WhatsApp groups and members' club washrooms.

[16:15] — Enterprise AI JVs: Palantir Meets Accenture Meets PE's Existential Crisis Anthropic and OpenAI both announced PE-backed JVs on May the 4th. Zero investor overlap. Andrew warns frontier labs risk becoming expensive consulting firms. Lomax: PE returns are compressed, LPs restless, AI is the new pitch to juice portfolio EBITDA. Everyone agrees this is IPO prep dressed in a consulting trench coat.

[23:15] — Mythos, Trump's AI Safety U-Turn, and Regulatory Capture Mythos is so good at hacking that Trump is drafting FDA-style pre-deployment AI reviews — having revoked Biden's AI EO on Day One. Andrew deploys the nuclear weapons analogy (drink!). Lomax offers the conspiracy theory: everyone in the administration has so much money in these companies that regulation now suits them. He then distances himself from that theory. Barely.

[31:30] — EU vs Meta/WhatsApp: Brussels Says Open Sesame, Zuck Says Pay Me Meta kicked rival chatbots out of WhatsApp, offered to let them back for a fee. Brussels rejected it. Mads lands the kill shot: open banking made UK payments vastly superior to America's. Sometimes the regulator is right. Potential fine: ~$16-20bn.

[39:27] — ASML: "No One Is Coming for Us" (Famous Last Words?) Fouquet went to Milken and said "come at me." Lomax notes China isn't replicating EUV — they're engineering around it with Huawei's CloudMatrix. Mads roasts Fouquet for lacking Jensen's paranoia. Lomax delivers the closer: founder mode vs. conference mode.

[46:33] — UK Fusion: Always 30 Years Away, Now Only 10 (Maybe) Sounds like a Bond villain's energy company: Gates money + American stellarator IP + UK magnets. 400MW by mid-2030s. Andrew: "magnets are not the reactor." Lomax: anemic growth, £120bn in interest payments, defence plan a year overdue. Building a fusion industry — or buying a franchise?

[56:09] — 🏆 Deal of the Week: SAP Acquires Prior Labs for €1bn 15-month-old startup. €9m seed. €1bn exit. The "GPT for spreadsheets." Possibly the fastest seed-to-exit in European venture history.

[57:30] — Week Ahead: Anthropic IPO board watch (they won't) · EU AI Act trilogue (~13 May) · Tencent Q1 earnings (13 May) · Fed transition — Powell's last day (15 May) · SpaceX S-1 (week of 18-22 May)

[59:46] — Dan's back next week. The nation breathes a sigh of relief.