With summer over, here’s what to look out for this autumn

Hello friends, and welcome to The Unsophisticated Investor! Brought to you by Mark, Head of UK Expansion & Operations at Shuttle!

If you want to invest alongside the VC funds who've backed breakout companies like Revolut, Monzo, Asana, JustEat, Bolt, Lets Get Checked, Loom, Runna, Charlotte Tilbury, Deel, Aircall, AngelList, Carta, TransferWise and many more, regardless of your knowledge, network or net worth, then sign-up today using the link below.

Now, let’s get into it 👇

The next few months are traditionally when European and US funding picks up. This autumn, investors are walking into a genuine tension: an AI ecosystem pricing an almost limitless addressable market while also carrying more off-balance sheet leverage and more physical grid constraints than at any point in this cycle.

 Equity analysts pricing the size of the market and credit analysts pricing the risk of the debt built to serve it are not disagreeing about the same facts. They are pricing different sides of a balance sheet that a war, a Fed chair navigating pressure from his own president, and an unresolved chip war are all pressing on at once. One number worth watching above the others this autumn is Oracle's credit default swap spread. It is the market's real time read on how the bear case and the bull case are being weighed against each other, and it tends to move before the equity market catches up.

In this week’s episode we look at what are some of the macro-forces to watch out for this autumn.

The financing question: TAM, capex, and the debt underneath it

Anthropic is reportedly preparing to tell IPO investors its total addressable market exceeds $30 trillion (£22 trillion), just ahead of the $28.5 trillion SpaceX pitched investors in its own prospectus in May. Both figures dwarf the previous IPO cycle: Uber cited a $6 trillion opportunity in 2019, and WeWork claimed $3 trillion before its collapse.

The scale here reflects aggressive valuation assumptions rather than a simple methodology shift; earlier tech cohorts defined TAM by specific, bounded service categories, while Anthropic's framing edges closer to the whole of human cognitive labour, priced as though a meaningful share of it could eventually run through one company's models.

Legendary valuations expert and NYU Professor, Aswath Damodaran, writing ahead of SpaceX's own offering, said this scale of TAM was reaching the edge of what is plausible. Anthropic's revenue reportedly grew from around $9 billion annualised at the end of 2025 to $65 billion by July 2026, and the IPO prospectus is expected to test whether gross margins and the scale of underlying capital expenditure can support a valuation built on a market sizing claim this large.

The strongest counter to bubble talk through most of 2024 was that hyperscalers were funding the AI build out from their own operating cash flow rather than debt, which gave the market a structural cushion against downside risk. That argument holds less well by the month. JPMorgan still projects cumulative global AI related capital expenditure reaching roughly £4 trillion through 2030, with data centre capex growth alone forecast to exceed 50% this year, so the scale of ambition hasn't changed. What has changed is how it's being funded.

Hyperscaler bond issuance averaged roughly £21 billion a year between 2020 and 2024. It reached approximately £89 billion in 2025 alone, and issuance in the first half of 2026 has already exceeded that full year figure. Oracle's five year credit default swaps, a reliable market proxy for AI debt sentiment, are trading at their highest level since 2009, and S&P downgraded Oracle's credit rating in July to one notch above junk status. A University of Chicago Booth analysis estimates a severe debt re-rating could reduce AI linked equity value by somewhere in the region of £7 to £10 trillion.

A good deal of this financing is structured off-balance sheet, which keeps aggregate risk exposure out of public filings. The £20 billion joint venture financing Meta's Hyperion data centre works this way: neither the campus nor its debt appears on Meta's own books, with funds managed by Blue Owl Capital holding the majority stake. Private credit intermediaries like Blue Owl are starting to show some strain of their own. Its business development company reported non-accrual loans at a five year high in August, and a recent fund merger crystallised a roughly 20% unrealised loss for the private fund investors being folded into it.

Meta's own lease commitments insulate its immediate financing from this, but the broader reliance on private credit intermediaries points to a fragility that off- balance sheet structuring was designed to obscure. None of this means the capex case collapses. It means the "this is cash funded" argument no longer holds in the way it did two years ago, and the autumn earnings season, when hyperscalers next report capex guidance, will be a useful test of where that argument meets actual numbers rather than forecasts.

The policy backdrop: a Fed under pressure, a UK budget, and a live EU AI Act

The Federal Reserve's upcoming decisions are colliding with macroeconomic friction from the Trump administration's own foreign policy. The war between the US, Israel, and Iran that began in February has disrupted oil exports through the Strait of Hormuz, pushed Brent crude above $100 a barrel for the first time since 2025, and led the OECD to warn that US headline inflation could reach as high as 4.2% this year.

This leaves the Fed in a tight spit. Hiking rates to anchor inflation risks compounding an economic contraction, while holding steady invites unanchored inflation expectations. The dilemma lands just as a new Fed chair settles in. Kevin Warsh, confirmed earlier this year following pressure from Trump to unseat Jerome Powell, has signalled openness to rate hikes amid persistent war driven inflation, which puts him in some tension with an administration that wants looser financial conditions from a governor it selected partly on that expectation. A September hold was the comfortable consensus through most of the summer, but Warsh's Jackson Hole speech in late August pushed markets toward pricing a hike as close as a 50/50 outcome.

Closer to home, the UK's Autumn Budget on October 28, and a major test for Burnham’s new Government, introduces its own uncertainty for domestic private markets. Speculation about changes to Capital Gains Tax and carried interest treatment could directly affect early-stage capital formation. If baseline CGT rises, it widens the relative value of statutory tax-sheltered schemes like EIS and SEIS, provided their core reliefs are left untouched, which is a rare case of a tax change potentially working in early-stage investors' favour.

On the regulatory side, the EU AI Act's Article 50 transparency obligations, covering disclosure for synthetic content and interactive models, became legally binding on August 2, even as enforcement for high-risk systems was pushed back to December 2027. The first wave of cross border enforcement actions from national regulators will likely set the practical bar for commercial deployments well before the broader 2027 mandates arrive.

Europe's concentration bet, and where liquidity is actually coming from

European venture funding rebounded 27% in the first half of 2026 to roughly £37 billion, and AI related companies took over 60% of total deployed capital, the first time a single sector has crossed that concentration threshold within a six month window; mega rounds above €100 million captured more than half of total deal value. Read one way, this is European private markets carrying a systemic overexposure to a single theme. Read another way, it is capital rationally concentrating where the only plausible asymmetric returns currently exist.

Three things will help settle which reading is closer to correct by year end: whether European LP commitments, which fell to a record low of €12 billion in 2025, show genuine signs of recovery; whether the diversification into defence tech and applied deep tech that everyone talks about actually shows up in closed deal data; and whether exit velocity ends up driven by trade sales and private equity buyouts rather than a real reopening of the IPO window.

Liquidity itself is splitting along an interesting line. The US venture secondary market hit an annualised £83 billion in Q1 2026, overtaking public listings as the primary liquidity route for the first time. That volume is heavily concentrated, though: the top 20 companies account for over 81% of all trading value, led by SpaceX, OpenAI, and Anthropic, so once these mega caps complete public listings and enter post IPO lockup periods, a large share of this secondary volume will temporarily disappear.

The market is responding by leaning harder into GP led secondary structures to bridge that gap. Global GP-led volume surpassed £89 billion in the first half of 2026, a 20% increase over the prior first half record, driven mostly by continuation funds.

The physical constraints: power and chips

Every valuation model pricing the expansion of AI compute rests on one physical assumption, that energy grids can scale delivery on schedule. Goldman Sachs Research estimates US data centre power demand will rise from 31 gigawatts in 2025 to 41 gigawatts this year and 66 gigawatts by 2027, with occupancy rates expected to exceed 95% before new transmission capacity comes online. This is pulling investment focus away from the asset light model layer and towards capital intensive infrastructure such as regulated utilities, grid expansion, and on-site generation.

It is also, usefully, where the most concrete good news of the year shows up. IRENA puts global utility scale solar at roughly $0.043 per kilowatt hour, against industrial grid electricity averaging around $0.22 in the EU, a gap wide enough to be changing where hyperscalers choose to build. Battery storage costs have fallen 45% in a single year to roughly $70 per kilowatt hour, which makes round the clock solar plus storage viable in a way it wasn't two years ago, and the US Energy Information Administration expects solar to be the single largest addition to American generation capacity in both 2026 and 2027, around 70 gigawatts, ahead of gas or any other source.

Solar doesn't solve the grid interconnection queue that is actually rationing new data centre capacity today, but of everything in this power story, it is the one part that has already been de-risked rather than merely promised.

The semiconductor supply chain sits at the same junction of sovereign industrial policy and private capital. Within China, Huawei is projected to capture roughly 60% of a domestic AI accelerator market worth between $30 billion and $35 billion, reflecting Beijing's $295 billion state backed push to build a semiconductor ecosystem insulated from Western export controls.

In the West, hyperscalers including Google, Amazon, Microsoft, and Meta are scaling proprietary in-house silicon programs at roughly 45% annual growth, compared with 16% growth for merchant GPU vendors, tolerating today's merchant GPU premiums mainly to meet near term compute obligations while building toward less dependence on outside chipmakers. Trade policy adds a further layer of unpredictability, with temporary export carve outs for capped hardware like Nvidia's H200 sitting alongside legislative pushes for broader bans on semiconductor manufacturing equipment.

Where the next big wins outside AI could come from

A few promising, if still early, signals are worth tracking alongside all of the above. Quantum computing has produced concrete milestones rather than roadmap slides over the past year: Google's Quantum Echoes algorithm demonstrated verifiable quantum advantage on a molecular structure problem, IonQ's medical device simulation outperformed classical high performance computing by 12%, and multiple groups have since shown logical error rates falling as quantum systems scale, the specific threshold the field has been waiting on since noisy, error prone devices proved too unreliable for production use.

Commercial traction hasn't caught up yet, and the gap between a lab demonstration and a paying customer is what will decide whether this autumn's milestones become 2027's revenue or 2027's excuses.

In deep tech, both of the more speculative upside cases share a reality that delivery timelines keep slipping. Isomorphic Labs' public clinical timeline for its AI designed drug pipeline has already slipped once, from an end of 2025 target to end of 2026, and whether an investigational new drug application actually gets filed this time will be the real test.

Fusion has missed its own delivery dates for decades, across both public and private money; the UK Infinity Fusion Consortium, formed in May between Tokamak Energy, US based Type One Energy, and AECOM, signals the same private capital phase transition already under way in the US fusion sector, but the reality is that capital intensive science reliably slips its own timelines.

What we’re watching out for

Four things will tell us a lot about where this cycle is really headed. First, Anthropic's IPO prospectus, which should give the clearest look yet at whether revenue growth (up more than sevenfold this year alone) can catch up with the scale of the ambition being pitched to investors.

Second, Oracle's credit spreads, which will show in real time whether the market's early-autumn nerves about AI financing ease or harden as more hyperscalers report capex guidance.

Third, solar and storage economics, which have already done more to change where data centres get built than almost anything else in the power story, and are still getting cheaper.

And fourth, whether Europe's record concentration of capital into AI turns out to be prescient rather than risky, a question the next few months of exit data and LP commitments should start to answer either way. None of these are foregone conclusions of course, but it will make this autumn an interesting one, and we’ll be closely watching it as it unfolds.

What we’ve been working on at Shuttle

• Final stages of testing product enhancements before going to production 🧑‍💻

• Advancing strategic partnerships for our UK launch 🚀

The Unsophisticated Investor is brought to you by Mark, Head of UK Expansion & Operations at Shuttle. Shuttle's mission is to break down the barriers to private markets and make wealth-building opportunities accessible beyond the ultra-wealthy elite. To do it, we're building Europe's private market infrastructure platform for running investment operations. By stripping away the legal friction and admin for deal managers, we make it easier than ever for high-performing opportunities to open up to a wider, modern network of investors.

Mark
Shuttle’s Head of UK Expansion & Operations