Blain’s Morning Porridge 23rd July 2026 – AI, Cheap Money, Tesla and Coffee.
“There is only one thing worse than having too little money, and that is having too much.”
It’s a wake up and smell the coffee sort of morning in the AI markets. AI is here to stay. Everyone is using it. Why are prices tanking? It’s about competition, adoption and data sovereignty. Maybe it’s time for a rethink? How did it get so huge so fast? That’s a question about overly cheap liquidity and speculation, which is why Tesla’s numbers are so interesting.
Yesterday Google confirmed the pace of its AI infrastructure spend will increase to $205 bln dollars next year. Second quarter income hit $112 bln, but it’s eaten into free-cash flow to the tune of negative $6bln while raising more debt, to “capitalise on the AI opportunity,” said the CFO. Exactly what is that opportunity?
The other hyperscalers will say much the same thing when they report on Thursday next week. Collectively the four hyperscalers are expected to spend upwards of three-quarters of a trillion dollars next year on data centres, networks, energy, water and chips. The rest of the AI sector reports next month, when we’ll learn more about adoption and the picks and shovels of the current investment Klondike. Collectively they are taking so much liquidity out the capital markets bond analysts believe the scale of the AI build-out underlies the current steepness of the US yield curve – the AI spend is “crowding out” the US government. It is the biggest private investment programme in global history.
But suddenly it feels a little wobbly. What comes next in AI?
As a bond market pessimistic pragmatist with a keen grasp of market history, I know hindsight is a wonderful thing. History shows most investment revolutions are defined by the same shaped curve – the exciting concept, the proof it works as the slope turns up, the steepening pace of adoption, brisk sales, and then, inevitably… the next new, new thing as the plateaux becomes a roller coaster back down. Suddenly the whole market is wondering exactly where AI is on the bubble curve.
Clue – we have seen this before. During the manic railway bubbles in the UK, Europe and the US in the 19th Century when hundreds of unbuild railways cost investors their shirts. We saw it during the Dot.com age when fibre optic cables were optimistically laid across the pond in the expectation of the parabolic growth of a new digital economy, the expected usage and revenues never materialised – they remained “unlit” as they rotted at the bottom of the Atlantic. There is nothing new in speculative investment puffery.
Underlying the massive amounts of capital flowing into equity and debt funding of the AI closed weight US models built by OpenAI, Anthropic, Google, and the amount of hyperscale infrastructure being built, was a very simple investment thesis. The plan was that US closed weight AI large language models would dominate the future monetisation of AI. But – no plan survives contact with competition and innovation.
The plan went something like this:
- The thematic narrative is simple – AI is a paradigm shift in the way all business will be conducted. Everyone will need and use it. Tick
- The adoption driver is the reduced costs to business and productivity gains that AI promises to deliver – irresistible. Tick
- The investment outlook is higher profits lead to generally higher market valuations. Maybe – depends on the cost.
- The Capex spending was driven by the expectation that enormous profits would flow to the owners of AI capacity who will rent it to businesses and individuals – what’s not to like about extractive returns? This approach misses the importance of data sovereignty. Er – maybe?
- AI will become a massively profitable utility owned by an oligopoly of hyperscalers, development labs and compute suppliers who will control and monetise data.
Simples… but it never is.
Now we are seeing push back across the curve. AI is great but it isn’t replacing jobs, it’s changing them. It isn’t cheap. Suddenly there are lots of places to buy the “picks and shovels” of the AI revolution. US tech has driven the global economy for 60 years now, but they no longer have a monopoly on invention and innovation. And, if the AI bubble pops, so will the whole market – let’s party like its 1999.
The reality is we now know the closed weight LLM models the US hyperscalers believed would give them market dominance are not the only option. Open-weight models – where the AI is a framework maintaining the users data sovereignty – have proved as good as the LLMs. Last week new Chinese open-weight SLM from Qwen and DeepSeek again stunned the Valley consensus with just how good – and how much cheaper – they are. Wow. US firms are suddenly finding out other nations also play tech.
Now we see the US AI collective mindset scrabbling to paint a new narrative that Open-Weight AIs are more “dangerous”, less regulated, and somehow… “unfair”? The Trump administration fears cheaper Chinese models will price out the US firms! The US trade rottweiler, Jamieson Greer, argues that “distillation” – the process by which AI models are now largely trained on other AI models – is effective IP theft (even as US firms have paid billions in fines for using pirate book sites to train models!)
Anthropic CEO Dario Amodei says open-weight models lack effective guard-rails. Yet they democratise the AI market by giving users sovereignty over their data. (The headlines this morning are all about an OpenAI closed-weight model that jumped the sandbox and started hacking competitors… but that’s ok because it was an American model devouring an American start-up!)
Yet, even in the US, we are now in a business world where everyone understands the value of data and costs, and how data sovereignty is now critical. As a result, the global economy has choices on how, what and from whom it buys its AI services, which is competition in action – and will keep it cheap and give business more options! Yay! That’s a good thing.
But it might mean they whole AI infrastructure build out goes the same way as the Metaverse… How has this happened? How have literally trillions of dollars been poured into the US AI bubble – on the basis a small number of US firms would dominate the market?
I’d start by looking at the US economy’s real “Unique Selling Proposition” (USP). That’s the depth of its capital markets. Since the 2008 global financial crisis ultra-low interest rates and QE effectively pumped liquidity into the US economy – but that was not invested in new plant, factories, or job creation. Nope. It was spent on stock buybacks and juicing the market – massive financial asset inflation caused by cheap money. That has left US investors the richest, the wealthiest and deepest pocketed moneymen in history.
And nothing creates speculative investment bubbles as fast as too much money…
At which point… let’s talk about Tesla and extrapolate what it’s results last night might actually mean. (For the record: my market credibility on Tesla is questionable. I sold 10 years ago and missed making a fortune. Win some. Lose some.)
I have always thought Tesla makes very fine cars. It is the most valuable car company on the planet. Yet Tesla is only the world’s 16th largest car producer selling 1.6mm cars last year compared to nearly 8 times as many Toyotas! Because of the increased competition in the EV space, Tesla is now cutting prices and has seen its extraordinary margin of 25% tumble in the last 4 years to 16% – which is still much better than the 5-7% margins the rest of the automotive sector makes. (They are more profitable because they sell more cars.) Tesla’s forward PE is around 170x compared to 10x for Toyota and 5x for VW.
Tesla is an extraordinary company. That’s because it effectively created the whole EV sector. It had first mover advantage, and it played its innovation ruthlessly. Yet, some stars shine so brightly they have limited lifespans. As I’ve written before; in 1903 the Wright Brothers invented flying. They had a monopoly on early aircraft. But competition, invention and innovation meant they were nowhere by 1914. They tried to reinvent themselves as engine makers, but the Brits and the French were way ahead of them.
Today, Tesla faces increased competition from China and established firms – all of whom are making very decent EVs. Tesla no longer leads. It is therefore pivoting to becoming the world’s foremost producer of Robotaxis and Robots. Yet… it has not sold a single Optimus robot or a Tesla Robotaxi. China is making and selling robots by the container load. Tesla is operating less than 150 Robotaxis in highly contained areas. Its energy storage business exists in market where Chinese producers are dumping cheap capacity on the global market.
So… what sustains the extraordinary value of Tesla? Its past as an innovator of a new market? Or its future in highly competitive commodity markets, where China and others like Waymo have a lead?
Go figure if it is worth what it is, and then figure what has sustained it? Cheap money, belief and hope – and hope in never a strategy.
Out of time and back to the day job…
Bill Blain
Author of the Morning Porridge
CEO Windshift Capital
Advisor – Spitfire Strategic Capital
Meanwhile, don’t forget about my new book:
You can read a review on the Society of Professional Economist’s website here.
The Battle For Hamble is a proper grown-up examination of how economies fail: a tale of Greedy Corporates, Bad Planning and Economic Illiteracy. It uses a wholly unnecessary Gravel Quarry in the middle of a prosperous village to illustrate the multiple failings and abdications of responsibility that have created Broken Britain. It’s about bureaucracy, money, exploitation and shareholders vs stakeholders. It’s about social injustice – asking why it’s ok to put 6000 jobs at risk so corporate bosses can reap bigger bonuses! Whether is the Hamble Quarry, HS2 or the abysmal state of UK armed forces due to bad procurement, read the book to understand how its broken process and bureaucratic indifference that is sinking Britain.
