Joe Hughes’ Post

There has been a lot of hype; understandably so over the past couple of years around AI, the adoption of, innovation, and implementation of and how that will shape the future of our world. One question that has always plagued me, is that of how do we supply the power needed for the rise of data centers and AI facilities? Does this align with goals of decarbonizing? Does it help or hurt? Should the leaders in the industry carve the path by enabling energy resources such as Nuclear and geothermal to play a role? Well, thanks to Sightline Climate (CTVC) for publishing this great article discussing this very topic. They focus in on SMR and a few of the key players including Google, Amazon, and Equinix that are acting to ensure sustainable growth. I would be curious to hear your thoughts. Will the continued growth in the AI space which imposes a 6x increase in energy demand for computing break the system? Or will it allow for the adoption of advanced energy resources such as geothermal, SMR and fusion? #renewableenergy #energy #data #AI #technology #infrastructure #sustainability #geothermal #nuclear #SMR #fusion #innovation Article ---> https://lnkd.in/gCZtY_vu

[2/2] AI = Energy. This has been our thesis all year. It was on display yet again this month when Microsoft signed a 20-year PPA with Constellation Energy to reopen the Three Miles Island nuclear power plant that was closed as recently as 2019. In doing so they are paying up to $100/mwh, plus a commitment to finance grid connections; all in they are paying a 200% premium to Amazon’s agreement with Talen Energy signed only last March. In a White House meeting this month, Sam Altman pleaded for subsidies to build 5 or more 5GW data centers to power AI across the US. The numbers keep getting bigger, with AI now prompting a global nuclear renaissance. AI requires 350TWh of new electricity capacity by 2030, prompting a 2.4x increase in global ex-China power generation growth. This power generation awakening is the core of our Cleantech allocation and was our top contributor for the month. Our portfolio allocation remains concentrated in Semis, Software and Cleantech. Our top contributing sector for Sept was Cleantech, with First Solar (utility scale solar), Fluence Energy (heavy duty stationary batteries), GE Vernova and Siemens Energy (global leaders in gas and wind turbines, grid tech and transformers), all up strongly; small caps Bloom Energy (hydrogen batteries) and Oklo (SMR micro nuclear) did not participate. Other top contributors were Meta, on AI boosting their advertising algos, and ServiceNow, confirming the strength of their AI Pro+ module. The underperformers were in Semis, notably Micron and KLA Corp, but we also suffered from a significant pullback in NovoNordisk and Lilly driven by tense rhetoric from US Senate hearings on the high prices of obesity drugs. For the third quarter, our best contributing sectors were Cleantech and Software Apps, while the underperformers were in Semis and Software Infrastructure. #disruptionfund #ai #energy

Bottleneck in the sense that there is less of any resource for generating electrical energy: https://lnkd.in/eDwZQtzD

The big 3 are starting to admit that power is the limiting factor to AI progress - the solution probably lies in compute efficiency, state space models, and linear scaling replacing the quadratic alternatives Zuckerberg admitted a couple of weeks ago that todays largest 150mw data centres ( equivalent power consumption of 4 million people) won’t be sufficient. To put this into context meta has the highest number of GPU’s at 300,000 and some of the biggest clusters at 24,000. The big 3 in AI are planning to acquire 10x this number H100 equivalents. (assuming 70% of the $130bn announced capex goes to hardware). 70 % going to hardware seems reasonable considering: (A) 75% of compute is spent on pre-training, (B) it’s still an accuracy arms race (C) llama 3 70b stopped training at 150 TR tokens and was still improving significantly (D) the next big usage will be on ‘ in context learning’ a much more compute intense usage . Data centre build is now restricted for power reasons in Ireland, Holland, and Austria amongst others . Latest UK new build proposals (UK is 3rd in the world by number) include small nuclear reactor proposals. The evaluators have no framework for approval though. I notice too that nuclear engineering graduates are at their lowest level for a decade…

Mark Zuckerberg: "Energy, not compute, will be the #1 bottleneck to AI progress" As Mark explains in the video clip: "I actually think before we hit [the capital bottleneck], you're going to run into energy constraints... because I just, I mean, I don't think anyone's built a gigawatt single training cluster yet." To put this into perspective, Mark highlights the scale of energy required for large AI models: "I guess put this in perspective, I think a gigawatt, it's like around the size of like a meaningful nuclear power plant, only going towards training a model." He also points out the regulatory challenges involved in scaling up energy infrastructure for AI: "Getting energy permitted is like a very heavily regulated government function, and if you're talking about building large new power plants or large build outs and then building transmission lines that cross other private or public land, that is just a heavily regulated thing, so you're talking about many years of lead time." While many companies are currently running AI clusters in the 50-150 megawatt range, scaling up to gigawatt-level clusters will be a significant challenge: "When you start getting into building a data center that's like 300 megawatts or 500 megawatts or a gigawatt, I just, I mean, just no one has built single gigawatt data center yet. So I think it will happen, right? I mean, this is only a matter of time, but it's, it's not going to be like next year." Mark believes that while the exponential progress in AI is likely to continue, hitting bottlenecks along the way is inevitable: "In general, you know, in history, you hit bottlenecks at certain points... I don't think that this is like something that can be quite as magical as just like, okay, you get a level of AI and you get a bunch of capital and you put it in, and then like, all of a sudden, the models are just going to kind of like, it just like, I think you do hit different bottlenecks along the way."

AI’s Power Demands Are a Ticking Time Bomb! 🤔 It’s no secret we've stumbled upon a formidable obstacle to our AI-powered future: the staggering energy consumption of our current models. Industry leaders are scrambling to come up with short-term answers to ensure they don’t miss the wave, with ambitious efforts like Microsoft’s reopening of nuclear reactors on Three Mile Island and Google working on “first-of-its-kind” geothermal projects. And while that plays out at the big kid’s table, a flock of new startups are building on the progress made in recent years, re-thinking the fundamentals to see if there are solutions that could serve as the long-term solution. One that doesn’t require hundreds of millions of dollars in infrastructure investment. https://lnkd.in/gFgEJQwY

We are at a crossroads as it relates to energy that is hard to reconcile with current western government policy. Here's the problem: AI seems to be real. We are at the start of a significant arms race. Between companies...between countries...and between civilizations (East v West). It doesn't seem like this will slow down anytime soon. What does this mean? There are two primary things that seem to matter in this race: 1) compute, and 2) energy. I'm not going to comment on compute...but it is clear that you can not advance on compute power and just do more of it and that gets you ahead. More energy = more advancement. The race is on, and that means that the requirement for more and more energy is getting ever greater...and this is not a linear thing...it is exponential. Google, Microsoft, and Amazon are already buying and building nuclear power plants. (https://lnkd.in/gvpxhSXm) This doesn't bode well for Canada (where I live), as our current government is doing everything it can to stand in the way of more energy. Some of the implications of this new AI world are becoming clearer...and everyone needs to start getting a handle on what will happen. If people are concerned about saving the planet (because carbon is supposedly bad), then nuclear is the only current scalable option. Our policies need to be changed to facilitate the approval of many new power plants. Otherwise, we are going to fall behind dramatically.

🔥 AI Data Centers Demand 1GW Power: Can Utilities Keep Up? The AI revolution is accelerating, with hyperscalers now projecting that future data centers will need up to 1GW of power to handle their immense computing loads. For context, that's the equivalent energy consumption of a small city! 🌆🔌 This surge in demand raises key challenges for the energy sector: How can utilities provide this power while maintaining grid stability? Can we scale up renewable energy to power these AI-driven facilities sustainably? What role will Battery Energy Storage Systems (BESS) play in balancing these loads and ensuring a reliable energy supply? At the intersection of AI and energy, bold innovations are critical. Utilities must rethink infrastructure strategies to meet these demands while ensuring a net-zero future. AI drives the next generation of data centers, and the energy sector must keep pace. 🔎 As we move forward, the question remains: How do we provide the massive power needed for AI without compromising sustainability? What will be the role Small Nuclear Reactors (SMR) will play? 💡 What do you think? #AI #DataCenters #Sustainability #EnergyInnovation #Utilities #PowerGrid #RenewableEnergy #1GWChallenge #GridResilience #BESS #NetZero

While the principle of additionality makes sense for clean hydrogen and the 45V tax credit (arguably, electrolytic hydrogen doesn't make much sense without the three pillars, but that's a separate conversation), imposing the concept on the AI-driven data center surge could do more harm than good. For a deep dive on this take, read my colleague Cy McGeady's latest piece.

AI data centers are driving demand for clean, 24/7 energy, pushing tech giants to invest heavily in renewable and nuclear power solutions. This surge in demand could accelerate the cost reduction of clean energy technologies due to their learning rates. In the long run, AI's energy needs may spur policy changes and advancements in clean energy infrastructure. #ai #technology https://lnkd.in/garvTktw