Cooper Westendarp

AI is burning Big Tech's free cash flow: The combined CapEx of Amazon, $AMZN, Alphabet, $GOOGL, Meta, $META, and Microsoft, $MSFT, rose +$36.9 billion in Q3 2026, to a record $201.9 billion. This figure has more than doubled over the last 12 months and is up +356% since Q1 2024. Big Tech has now spent $1.6 trillion on CapEx over the last 3 years. Meanwhile, the combined free cash flow of these companies fell -$33.1 billion last quarter, to -$26.4 billion, its lowest on record. Free cash flow among these firms has dropped -$88.2 billion over the last year. Only Microsoft reported positive free cash flow in Q3. Looking ahead, these 4 companies plan to spend another $232 billion on CapEx in Q4, potentially lifting their combined CapEx to a record $740 billion in 2026. Big Tech is undertaking one of the largest infrastructure buildouts in history.

From the WSJ, and further to earlier posts: "For years, the companies behind America’s artificial-intelligence boom have raised hundreds of billions of dollars with ease. Wall Street eagerly ponied up capital and asked few questions. Now, the balance of power is shifting. Lenders and investors are getting choosier about where to place their bets and are demanding tougher terms, signaling a new phase in the AI build-out." #economy #tech #ai @WSJ #ArtificialIntelligence #markets

And still bullish $NEAR... That deep dive that @RaoulGMI refers to was deep in the bear market in Q2 bases purely on the growth in intents and capital discipline ( lower emissions and buybacks). We are now seeing the start of the monetisation of the next revenue pillar, Private AI inference.

It begins...

And what binds power? • land • permits • grid access • transformers and switchgear • noise • clean air • supply chain • fuel security • casting machines • suitable large lathes for turning shafts • residential priority • community political power All the best sites are now taken, and every year the marginal development site gets worse and worse.

As someone with a portfolio of energized land assets here is what I am seeing: 1. Neoclouds are disorganized and poorly capitalized. Some are definitely better than others however. 2. Hyperscalers are intractable, bureaucratic organizations that take 3 months to do an NDA and want 18mo escrow periods. We built our portfolio by competing against these terms because of our speed and simplicity. The more future assets orgs like ours win, the bigger the problem for the hyperscalers downstream (see 3). 3. There are delays in the existing pipeline that will catch people off guard in 2027. Hyperacalers will (obviously) react by focusing on their own internal compute needs over any third party. If the third party is a model company, they will tell them to go through their model router where they will also take a non-trivial vig. Them telling Ant they have no compute for them (rumored in Ant’s S-1) if true is a huge tell that this is already happening. All of this is to say that it does look like there will be a lot of partially stranded compute and compute sitting in inventory in 2027 than is currently forecasted. This is bought and paid for but effectively underearning (below total cost of capital) or collecting dust. We see this as true for everyone - model companies, hyperscalers, neoclouds except ONE (you can guess which one). This dislocation will benefit two kinds of orgs: the non-bureaucratic big scalers who can build industrial scale quickly and the nimble, well capitalized players feeding them. 2027 will be a turbulent year because it will clear out a bunch of wrong/bad bets. It will also be a consolidation year for a few.

Palantir CEO Alex Karp on the new AI security reality: “With these new models, it’s going to be more guerrilla war… how quickly can you mitigate and how quickly can you shoot back.” “The models are able to expose security flaws at a scale that we’re not really equipped to deal with. That’s going to be a societal issue for every single organization, government or not.” “The only unique thing we’re doing… is how can you patch the penetration vectors that were unknown quicker than they can be exploited?” The age of static defenses is over. AI is turning every network into a live, constantly shifting battlefield. Most orgs will be permanently on the back foot. Palantir bet is that the only winning move is speed. Finding and sealing the holes faster than the models can open them.

We are at Security Armageddon now. A large part of internet is affect in some way today, lots of older software vulnerabilities are coming out and through those more software is getting exploited. PHP systems like Wordpress power a big part of content websites. Malicious TLS certificates especially affect basic services and older codebases. Plus node.js, plus LiteLLM, etc etc. Uncensored models are becoming a key to enable security defenses to keep up. Formal verification allows to improve on how things are verified but still requires smart models to translate between systems and make sure formalizations are correct. We also really need to transition to models that involve cryptography where right now API keys and access tokens are used. Hosting platforms that can secure the software even from platform itself. Meanwhile, update, patch, rotate.

Hard to believe, but the hottest market on earth, AI compute, has no real exchange. Deals are OTC, by phone and PDF, like oil in 1983. It cannot go on an order book: GPUs are heterogeneous like oil, an H100 on InfiniBand in Virginia is not an H100 in Jakarta, and perishable like electricity, an idle hour at 3am is gone forever. Nothing with both properties has ever traded cleanly. So H200 hours cost what they did at launch and the whole market clears over Signal. This is a NEAR Intents shaped problem. You cannot post an order for a thing with no standard grade, but you can post a need: a thousand GPUs of roughly this shape, this region, these weeks. Solvers who have it, or can stitch it from five clouds, come back with a price. The biggest market without an exchange will get one, and it will look like near:native Intents, not Nasdaq.

Week 3 of NRR: Another 1.4M $NEAR added since my last post. ~17.1M now sitting in the Bitwise NEAR ETF. TradFi keeps accumulating.

1/3 HSP-029: Establish a 24-Month Issuance Ramp to a 1.6% Target Rate of NEAR Token Emissions — Vote now! This proposal aims to gradually reduce $NEAR's annual token issuance from 2.5% to 1.6% over 24 months, helping reduce token dilution and strengthen long-term tokenomics.