Akash Trips

Much of Dwarkesh's argument hinges on this statment which *was* accurate but will be increasingly inaccurate on a go forward basis imo:    “American labs port across accelerators constantly. Anthropic's models are run on GPUs, they're run on Trainium, they're run on TPUs. There are so many things you can do, from distilling to a model that's well fit for your chips.”   As system level architectures diverge (torus vs. switched scale-up topologies, memory hierarchies, networking primitives), true portability is eroding. The Mi300 and Mi325 had roughly the same scale-up domain size as Hopper while Blackwell’s scale-up domain is 9x larger than the Mi355 scale-up domain, etc. Many frontier models are now being explicitly co-designed for inference on specific hardware like GB300 racks. Codex on Cerebras is another example. Those models run less efficiently on other systems and the performance differentials will only widen. A model that runs well on Google’s torus topology will run less efficiently on Nvidia’s switched scale-up topology and vice versa - the data traffic is fundamentally different as a byproduct of the models being parallelized across the different topologies. Google’s internal teams - and increasingly the Anthropic teams as they become the most important customer of almost every cloud - have the luxury of operating across the stack (models, chips, networking) - but that is not the case for the rest of the market and other prospective users. Anthropic is the exception, not the rule. To wit, Anthropic and Google allegedly have a mutual understanding where Anthropic can hire the TPU engineers they need every year to ensure that they can continue to get the most out of the TPU. Given the overwhelming importance of cost per token to the economics of the labs, models will be run where they run best. Most extremely large MoE models will run best on GB300s given the importance of having a switched scale-up network like NVLink for MoE inference. When training was the dominant cost for labs and power was broadly available, labs were optimizing to minimize capex dollars. Model portability was a way to create leverage over suppliers. I think that drove a lot of the focus on portability. Today, inference costs as measured by tokens per watt per dollar are everything. Inference is way more important than training costs (inference is effectively now part of training via RL). Labs are therefore now optimizing for inference. This means increasing co-design and higher go-forward switching costs for individual models between systems. I do think this explains why Anthropic and Nvidia came together: Anthropic needed Blackwells and Rubins to inference at least *some* of their models economically. And Mythos might just end up being released coincident with the availability of Rubins for inference. TLDR: as labs shift their focus from training to inference, the costs of portability and the upside of co-design to maximize tokens per watt per dollar both rise. Portability is likely to begin decreasing as a result.   I think what I might have respectfully added to Jensen’s answer is that systems evolve under local selective pressures. The evolutionary pressure in America is a shortage of watts so it makes sense for Nvidia to optimize, as an American company, for power efficiency and tokens per watt and stay on copper as long as possible. China has a surfeit of watts. Chinese AI systems are already taking advantage of this with the Huawei Cloudmatrix 384 and Atlas SuperPoD having an optical scale-up domain that is much larger than anything offered by Nvidia today at the cost of *much* higher power consumption and much lower tokens per watt. The networking primitives for this Huawei system are very different than those for Nvidia’s systems and a model that runs well on Nvidia will not run well on that system and vice versa. This means that if a Chinese ecosystem gets momentum, Chinese models might stop running well on American hardware. And when Chinese models run best on American hardware, America is in a better position as this gives America a degree of leverage and control over Chinese AI that it risks losing to an all-Chinese alternative ecosystem.   This architectural fork makes porting and distillation less effective and strengthens the pro-American national security case for selling China deprecated GPUs imo. Also I will attest that I did not wake up a loser this morning.

I just killed my bullish thesis on $IREN and out of every position. In this video I walk through the exact trigger that flipped it into a potential bear cycle, and why I see a possible 40% downside over the next few months. I hope I’m wrong. Here’s the full breakdown 👇

After six months of thorough due diligence on $NUAI—reviewing filings and engaging directly with management—I went all-in on this hidden gem. The company's behind-the-meter strategy is the way forward in today's power- and compute-constrained world. If you believe in management's plan to sign an investment-grade hyperscaler in the coming months, there is little doubt that this company is worth $3–5 billion. Yet it trades as a highly speculative small cap, currently valued at around $255M. Therein lies the opportunity. Most on X know me as a long-time bull on $IREN, a company I've been all-in on for the better part of 2.5 years. So this portfolio shift isn't one I take lightly. For the doubters, $IREN is, pound-for-pound, the best business in the AI HPC space today. In fact, if Mr. Market hadn't given me such a wonderful entry point into $NUAI, I would still be all-in on $IREN. The investment thesis: While both companies provide energy-intensive AI infrastructure, $NUAI's primary thesis is its valuation—materially below peers pursuing comparable opportunities. Few companies, if any, focus exclusively on behind-the-meter power delivery, sourcing directly at generation sites to achieve lower costs, higher efficiency, and independence from grid interconnection delays. This model makes the company highly attractive to prospective tenants amid rising demand for reliable, scalable power in AI compute. Like in most power markets today, the past 18 months in $ERCOT have seen substantial challenges for large-load grid interconnections. Datacenter-driven demand expanded ERCOT's queue from approximately 63 GW at end-2024 to over 230 GW by late 2025. The prior sequential study process resulted in repeated restudies, modeling inconsistencies, extended timelines, and uncertainty. Disguised as an attempt to help with those problems, ERCOT’s new batch process plans to group qualifying large loads for coordinated transmission studies to improve efficiency, consistency, and system reliability. For grid-dependent projects, this adds procedural layers and timing considerations during a period of high demand. Some would say this is a major step back for allocating capital to ERCOT. Meanwhile, $NUAI's behind-the-meter approach circumvents these grid-related hurdles—no queue participation, no restudies, no batch constraints. Power is provided directly to tenants through dedicated on-site power plants. $NUAI's first 3 GW will use gas-fired technologies, giving the company and its tenants unparalleled speed to market. $NUAI management expects to announce the first phase of TCDC (Texas Critical Data Centers) within the coming months. From there, the odds of ramping up quickly to 1 GW remain high. Management has stated repeatedly that they are in advanced discussions with multiple hyperscalers. It's worth noting that previously guided timelines have slipped, but the company has announced partnerships that shore up uncertainty around the desired outcome. Assuming management can get $NUAI to the promised land, we should see its market cap shoot up to $3–5 billion in very short order, finally bringing it in line with peers. That's a 12–20x in a matter of months. If management continues to execute on its NM site and others, NUAI could conceivably reach a valuation of $20bn—almost 100x from here—but we'll leave that for another day. The downside: What if the company isn't able to deliver on a hyperscaler tenant? Like any good investor, I am laser-focused on downside protection, and that's where $NUAI is a no-brainer. TCDC is a 438-acre site fit for a 1 GW AI datacenter. Located across the street from two large power plants owned by the likes of $VST and $FANG, and with three different natural gas pipelines serving the site, TCDC is the equivalent of what is called a “powered land” site. These days, powered land is valued at approximately $500,000 per MW, implying roughly $500 million in embedded land/power value for TCDC. With $NUAI's market cap hovering around $255 million, the land value alone exceeds the enterprise value, with the operations, partnerships, and development pipeline—including its 3,500-acre New Mexico site—providing additional upside. I describe the range of outcomes as “heads I win big, tails I still win.” The noise: The past few months have seen the company come under attack on several fronts. The most prominent is a civil complaint from the New Mexico Attorney General involving allegations related to oil-well responsibilities transferred through entities. The company has described the claims as unfounded, with no prior state engagement, and intends to defend vigorously. Maximum potential settlement exposure, based on disclosures and precedents, remains below $10 million. Having diligenced this with my own network of seasoned litigation attorneys, this is a nothingburger. Portfolio sizing: Concentrated sizing suits my investment approach: when extensive research identifies strong fundamentals, a clear competitive advantage, and attractive risk-adjusted upside, I allocate accordingly. I’ve taken over a decade to get here, so this is not a good idea for the vast majority of investors. To be clear, I don't go all-in often, but I do run a concentrated portfolio of 2–4 stocks. The past few years have been very different because the big winners have been sitting in plain sight. The first all-in idea in the past few years was $VST, a power production play, followed by $IREN, a power-meets-AI compute play. At their core, both are infrastructure companies within my circle of competence. $NUAI is just the next evolution of this very same AI infrastructure thesis. It's not every day you see 10-to-100 baggers sitting there in plain sight. And yet, here we are. $NUAI presents a highly asymmetric opportunity in energy-constrained AI infrastructure, with the valuation deeply disconnected from the company's intrinsic value. Do your own research—this involves risks inherent to early-stage development and sector dynamics. Not investment advice; simply outlining my allocation rationale.