$ANT can cancel its pricey compute deal with $SPCX one part on Jan 27 and the other on March 27. One should not model higher compute prices like this as the default long-term, especially considering $ANT has the biggest negotiating power being the biggest buyer. It seems like, on a sell-side call, $ANT said that they will NOT extend the $SPCX deal beyond the middle of next year. This should get a lot of attention and should put additional pressure on neoclouds, as compute prices have a limit.

Former $GOOGL employee on why data center capacity and electricity are the real bottlenecks, not GPU production itself, and how unreliable some of the neoclouds are ($AMZN, $GOOGL, $MSFT, $NVDA): 1. According to the expert, open-source models are closing the gap with frontier models faster than expected, with customers increasingly replacing Anthropic and OpenAI models in multi-agent architectures with open-source alternatives starting from Q2 of this year. The shift started with simple tasks like tool calling and MCP, but has moved up the stack to low-level orchestration. 2. The expert identifies capacity as the biggest constraint at the moment, with certain companies' traffic growing 5x while GPU availability, particularly in the U.S. where latency requirements are tightest, simply cannot keep up. Sourcing capacity overseas in Asia or Australia is technically possible but adds latency that most customers will not accept. 3. Hyperscalers offer reliability but are expensive, require multi-year commitments of three or more years, and have limited availability of the latest hardware, with customers already chasing B300s. AI-native companies tend to gravitate toward neoclouds due to hyperscaler pricing and contract rigidity, but neoclouds come with their own reliability issues, with GPUs dropping off unexpectedly and requiring engineering resources to manage fleet stability around the clock. 4. The third tier consists of smaller independent operators who have invested in GPU racks, seeing the opportunity to lease out capacity and recoup costs within a year or two, with the remainder being pure profit. The expert describes these as essentially unusable in practice, with operators who do nothing more than find land, install racks, connect electricity, and set up cooling, with no real cloud management, maintenance, or operational expertise behind them. 5. The expert expects overall GPU availability to increase meaningfully over the next year or two, with $NVDA continuing to ramp production. The real bottleneck is not GPUs themselves but data center capacity and electricity in the U.S. On the supply side, the expert also sees players $SPCX beginning to lease out spare capacity to the market, with more likely to follow. 6. The expert notes that U.S.-based neoclouds typically require 20-30% prepayment upfront, with the remainder paid monthly or quarterly over the contract term, and rebates negotiated at the end for any reliability incidents like traffic loss. APAC-based providers operate differently, with no prepayment required at all. found on @AlphaSenseInc

What a lot of folks don’t understand is that the more users $META Muse is getting at the scale it is already (+5M downloads) the more feedback it is gathering to improve the product the bigger the moat is. The first mover advantage could be significant here, since coding (making the product better, fixing bugs) is actually something which AI can solve in a loop much faster than what we had in the pre-AI era of apps (feedback->fix/improve->repeat can now be a matter of minutes and used to be months).

$META trades at a fwd P/E of 22.7 yet: - The core business is growing 30% & becoming more valuable in the AI world (reaching a human via ads will become a premium), - $META has a strong head start and lead in the personal AI agent race with Muse (Muse growing at a pace as one of the fastest consumer apps in history), - $META having 3.5-4 GW of compute capacity (the most sought-after asset in the world right now)

I published my article on the Agent Economy and how I think personal agents like $META Muse will reshape industries and business models. 1. Explaining how AI agents browse and connect to the world 2. Revenue layers at risk (friction, attention, search arbitrage, etc.) 3. Where we might see a positive boost in usage/revenue from the use of personal AI agents 4. Shared a list of over 30 public companies I analyzed and rated according to the risk of full or partial business model disruption (including my view on $AMZN and $GOOGL) 5. Explained how incumbents can play it and defend/keep their moats I had a great time researching, analyzing, and putting this piece together; I hope you will too.

Wow, OpenAI making dots the $META Muse competitor only available to Pro plans not Plus, Go or Free and looking at the features it doesn’t look anything better than Muse. $META really has the road open now to take the consumer AI agent market.

It does seem like OpenAI is out of compute. I wonder how they will be able to provide compute for both a personal AI assistant & support their work use cases. We will see what they release today, but the $META Muse competitor doesn't seem like it will be free or scalable with this compute.

The personal AI agent race will be decided between $META, OpenAI and $GOOGL. The personal AI assistant market at the end comes down to who can serve the agent with the model that is good enough for the lowest cost. For that you actually need to have your own model or open-source always avaliable. You also need enormous AI compute and traditional infrastructure. That is why I think it comes down to $META, $GOOGL and OAI, as $AAPL doesn’t have the data center footprint or their own models. In terms of the other three $META seems best positioned with the first mover advantage, strongest distribution and more compute avaliable as both $GOOGL and OAI have a strong enterprise offering to support as well. If $META can scale to hundreds of millions of users and keep the product free it is likely game-over for others.

Investors still vastly underestimate the hyperscaler data center layer ($AMZN, $MSFT, $GOOGL) as open models are taking more share. Some valuable insights from Vercel CEL @rauchg: »When we started the AI gateway, it was all OpenAI, and then a lot of it was Gemini, and then a lot of it was Anthropic. And those three have consistently been losing ground to open models.« »Open models and faster models seem to be the new trend.«

The AI agent market is not zero sum market. I believe each of us will want at least two seperate AI assistants in their life. One being a personal one covering their personal life and the other being work-related. $META and $MSFT are well positioned and made the first moves. The reason why we will want to seperate those two roles is because each agent must prioritize different goals and objectives: 1. A personal assistant is meant to optimize and take in account your life goals and preferences. It should also be high in privacy and not meant to be observed, audited or transfered. 2. A work agent will probably be a owned and shared with your employer. Once you leave a company or change jobs the agent will get deprovisioned and the memory will stay with the employer (that is why you don't want any of your personal stuff in there). The employer wants the agent to be observed, audited all the time. Because of privacy and safety your employeer will also not approve of all the integrations and connectors that your personal agent would want because it poses a risk to company data and cyber risk. The way I believe we will have these agents is that the personal and the work agent will communicate on matters that they need to like calendar scheduling or other tasks, but each keep their own data seperate and siloted from each other. I made an image of an example I made for $META's Muse potentially communicating with $MSFT Autopilot. For the personal assistant having the best IQ model doesn't matter its more a skillset similar to human assistants: knowing the boss's preferences, holding the keys to the calendar and inbox, knowing when to interrupt and when not to, being fast, being discreet, and being reliably available. $META is in my view positioned great for this as it is the closest to the current system of "knowing your preferences" from the social algo that they have on us. For the work agent it does seem like companies will want to be independent and not reliable on only one model vendor so an orchestration layer (between models & model companies) is almost needed here, especially as the workers tasks can change over time and different models are good at different things.. Here the companies that are very well positioned are the hyperscalers and especially $MSFT because they have many of the integrations that the assistants needs (human tools) + they have already established relationships and trust from companies. $MSFT again doesn't need to own their own frontier model they just need to be the best router and keeper of context and data (harness). The point I wanted to get across is that the AI assistant market is not there is only going to be one AI assistant that wins, as we will soon get a flurry of different AI assistants.

$META will raise funds & increase CapEx significantly as Zuck just said on a podcast that he thinks the way to get to ASI is to brute force it with more compute. He said they want to get there and that compute is the way. He is signaling you exactly what $META is going to do.

It would be very funny if, in the AI era, at the end, $META wins the personal AI agent market and $MSFT wins the enterprise AI agent market, in a way that you have your $META Muse and $MSFT's Autopilot communicating across different use cases (private and work). Distribution was key in the pre-AI era and, as it turns out, is key also in the AI era.

Now that raw model performance isnt the biggest driver of usage anymore $MSFT has the perfect chance to be the “Muse” of enterprise as they can put together a product with the best harness and integrations. While $META has the strongest distribution with consumers, $MSFT has it with enterprises.

Less than 2 months, $META up +33%: "you can see that in the next few years the company will not just be an ad business but will have at least two massive new revenue verticals: transaction revenue on e-commerce and AI agent revenue for businesses & creators through its platform like WA"

It will be hard for anyone to compete with $META Muse. You either compete on the product, distribution or price. $META is really strong in all three categories: - The product reviews for Muse are 40k ratings with a 4.88 score (truly rare) telling you users love the product. - In terms of distribution, $META owns the biggest ecosystem by far. - In terms of price $META has the most spare infra capacity + enough profits from its core to subsidize if needed.

If $META dominates the personal AI agent market, it becomes the ultimate stack for businesses trying to reach consumers. Not only does $META then own top-of-funnel discovery surfaces with FB, IG, and WA, but it also adds a high-intent commerce gatekeeper (the agent), which means it owns the full stack and can do full attribution. When valuing the company, you must add the value achieved by owning and offering both discovery and high-intent, bottom-of-funnel marketing surfaces. In essence, this is the holy grail of social and commerce.

I published an article on what it means for every part of the AI stack if frontier labs (Anthropic, OpenAI) start to slow down their development. - What is scaring the AI labs & RSI - Implications on hyperscalers ($AMZN, $MSFT, $GOOGL), semis & app layer

Anthropic when it comes to pacing of AI development wants AI labs to allocate more compute to safety. They also shared that they are spending around 6% of their AI R&D compute on safety.