David Riester’s Post

Finally got around to digesting the Goldman Sachs paper on #AI from a couple months back (https://lnkd.in/gZutQ8jN). Its good. Many of it's predictions have borne out since it's publication. The same could be said of Sequoia Capital's (David Cahn's) pontifications from the same week back in early summer. Few (zero?) financial/technology bubbles have more far reaching long-term ramifications than the #AI variant. Usually when we talk about bubbles popping, our minds go to silly websites (Flooz! Pets.com, etc), Webvan, McMansions in foreclosure, or Dutch tulips. While the housing bubble touched almost everyone in the world - some very tangibly and deeply - society has mostly recovered, and markets undoubtedly have. However, the secondary/tertiary #power, #climate, #energytransition, #landuse impacts of the #AI bubble are enduring and mostly irreversible. I discussed this in a white paper about developing power into the #datacenter demand theme recently (https://lnkd.in/g_X-Gag2). So, while I'm of the same mind and rather impressed with the financial/investing predictions of Goldman and Sequoia, I worry that their investor lens inadvertently encourages us to consider the issue as we might any other boom/bust cycle. A sort of "be careful not to get burned financially here; there's irrational exuberance and the numbers don't quite add up". Which isn't wrong in the slightest, but - because of those secondary and tertiary impacts - undersells the need for discerning, farsighted, multi-discipline trepidation that extends way beyond one's investment portfolio or corporate strategy. Overshooting this will leave deep scars on our country and the world. Yet... i) AI is NOT a bunch of baloney, ii) we SHOULD be building a lot of infrastructure to support the emerging and exciting sector, iii) businesses and people SHOULD be excited to explore genuinely accretive use cases, iv) there is unquestionably a lot of money to be made in the space, and v) #energy #power developers SHOULD be rising to meet the load needs It is a matter of degrees, and pacing These things can all be true at the same time. We have trouble with that, psychologically. We have trouble finding a sort of equanimity on matters like this. There doesn't need to be "camps". We need sober minds looking at the question of how AI should be unfurled upon the world in a way that does more good than harm. That probably means moving a little more slowly than we have been. It means batting back the FOMO. It means shifting our vision to long-term gains instead of short term windfalls. It means building out the infrastructure at an "aerobic" pace - one where data centers are powered mostly by clean power - not an "anaerobic" pace that leaves us with the lactic acid of a bunch of hastily approved/constructed thermal power.

Every year #re+ offers a chance to tap into the zeitgeist of the #energytransition for a couple days. Here’s what stuck out this year: - Firms (all categories) that have prioritized focus over expansion seem to be thriving. I enjoy playing a little game wherein after someone states a desire to grow, I perfunctorily ask - "Why?". You might rightly infer that the answers rarely satisfy. There are fine reasons to want to grow, of course, but usually the most honest response to that question is "inertia", or "cause' other shops do", or "to justify our bloated SG&A we need to expand our footprint" (which, of course, leads to the tail-chase of more hires, not increased leverage of the existing team. - A related phenomenon, which – combined with expansion plans – is causing problems: developers that want to "become an IPP". It’s a tale as old as #renewableenergy, yet somehow getting worse not better. Here again, asking "why?" is fun, and here again there are rational reasons, but those are infrequently offered in response. Today's market is full of developers sitting at a dangerous confluence of circumstances: i) desire to expand strategy, + ii) the associated bloated SG&A, + iii) the desire to retain operating projects (therein forgoing liquidity), + iv) increasing #interconnection time horizons and deposit requirements. The math is simple - No one can navigate that without a significant injection of capital. - …and the universe of investors able to close the deals that would inject that capital is far thinner than people perceive. Note that I used the words “able to close” not “interested in," between which lies a huge gap causing dozens of dead deals that leave developers in a pickle. - Underneath all this is a market bursting at the seams with good projects, and a staggeringly deep, rich talent pool. - And the #IRA rollout is going very well - especially #onshoring and #transferability - Put all that together, and the obvious question is “when will the denial give way to consolidation?” I don't know, but I'm confident it will come in the form of company valuations equal to the sum of their projects' asset value (i.e. no enterprise value). Which, for most developers, would be deeply disappointing.  I expect that many of the leadership teams who, finding themselves cornered as such, will look back and wish they'd prioritized focus over growth, liquidity over "becoming an IPP", simplicity over complexity, and right-sized SG&A instead of letting SG&A levels dictate strategy choices. A little self-honesty, pain, and humility now can avoid massive disappointment down the road. Like a sailboat poised to ride the Tradewinds straight across the pacific, the marginal "cost" of deviating course to a pacific island to try and buy an engine is too high. Rather, the right play is to take care of the vessel & sail to ensure you remain in the game, positioned to ride that tailwind to glory.

None of us need another article about #datacenter #electricityload growth and the implications for the #energy mix, but perhaps more perspective on how (or how NOT) to develop #power facilities (#solar, #wind, #bess) into this theme is welcome? As usual, I've attempted to discover some pithiness in my toolset, and failed miserably. But there's a lot to cover here, so dig in. If this sort of content is up your alley, the Segue Sustainable Infrastructure "Insights" page (https://lnkd.in/ggZx-bVh) has a reservoir of it. This article will be posted there shortly. Thanks to Anne Marie Denman for the nudge to cover this turf, David Berry of Cloverleaf Infrastructure for the perspective, and Dillon Stambler of Creative Renewable Solutions, LLC. for the same.

I took last week with the family on the big island. There was a centrally located and very chatty hot tub at our spot. Straight away I could tell this Jacuzzi-crucible held the potential to be a fascinating sample set of America’s well-off. So I kept a little journal of the 9 conversations I had: 1. A highly educated, middle-aged man who struck me as very intelligent and endowed with skills/experience that could be applied to numerous fields; he is a partner at a bay area #crypto hedge fund 2. Two 20 year old's who had recently dropped out of college to “enjoy life more and stuff”. They were at this resort drinking daiquiris on their parents’ dime, declaring that “this is what [they] want to do for their life.” 3. A very well-bronzed man who sources fine art for wealthy individual buyers, with a strategy of procuring pieces from financially troubled museums 4. A woman who sells advertising space for formula 1 television broadcasts 5. A couple on their honeymoon, one “between careers”, the other a master carpenter moonlighting towards a Retail Investment Advisor qualification because "carpentry can't support our family, even with the financial support we get from our parents".   6. An "emerging influencer in the cosmetics space"   7. The "CIO" of a "private investment firm", which, upon further investigation, turns out to his own family's family office (his "job" is investing wealth generated by his ancestors)   8. A real estate agent from Orange County   9. Co-founder of an AI-oriented startup that that just closed their series A at a valuation between $50-75m; this co-founder was on day 2 of an 11 day vacation    Look, I know I'm being offputtingly judgmental and holier-than-thou. The point isn’t to take shots at any of those specific individuals or their life choices...rather I'm reacting to the impression left by the whole of it. My hot tub convos painted a picture of a society/economy at risk of failure due to the hollowing out of substantive value creation. There's nothing in all that stuff. No substance. Lot's of parasitism ("rents"). As Ray Dalio might frame it, there's a sort of decadence-wavering-on-grotesque here.   And I'm not having that reaction through any political lens. I don't think this is about politics, per se, but rather human nature, sociology, and long time-horizon economic cycles. Much of American affluence stands on the shoulders of i) value created generations ago, ii) the ongoing printing of more money, and iii) the luxury of having the world's #reservecurrency. Those things are all ephemeral... promised/owed to no one. I just wish more of us felt and responded to that reality, such that a movement toward more purposeful application of our finite #humancapital might help our geriatric empire rage against the dying of the light.  Thanks for enduring my preachiness and pessimism. FWIW I've plenty to criticize about the paths I've chosen to take myself. (image is Dalios from "The Changing World Order")

It's very fun to be at a point where Segue can pull in folks like Neerav who have considerable depth of directly-relevant experience, and an intuitive sense of what we do at Segue. Not to mention a fantastic network in the industry. Speaking of which...if you're an old friend/associate with some projects that need development capital support, Neerav is ready for your call :)

Great gag gift from Jennifer Williams at Live Oak Bank. We like to have a good laugh about how Segue has taken a perfectly good engineer (I'm being humble on Joe Song's behalf) and ruthlessly contorted him into an investor. Joe likes to say things like "I'm a big promissory note guy these days", and even has a suit jacket now. To all his many #engineer pals, rest assured his sellout is half-hearted. He can still dive into a PVsyst, a cost buildup, or a Gannt chart with the best of them. Like Marlo Stanfield at the end of the Wire, he takes it back to the corners now and again, just to knock some emotion loose and make sure he's still got it.

I always feel like a buzzkill when making this point, but it seems really important and not commensurately "in the discussion." We're all excited about the potential, and even some of the existing applications, of #ai. I'm considerably less into it than most, but obviously still see the potential. A considerable investment of i) energy, ii) human capital, iii) $ capital is warranted. But we're not talking much about exactly *how much* (of each of those) is appropriate. Where does the cost/benefit tip? Or, said another way, where are the marginal gains of the next AI #GPU center less than the marginal cost (including intangible costs). Using extremes to illustrate the point, does anyone think we should pull out all the stops, increase our global power output 2x hell-or-high-water, allocate 100% of that marginal #power/ #energy to #artificalintelligence, even if the actual and opportunity costs (real estate, time, energy, raw materials, money) were outrageously high? I don't think any serious person can answer yes to that. So there's a "line" somewhere. Finding it is not a purely objective exercise, but we need to acknowledge the fact this line exists, and make decisions about AI infrastructure within a framework that does the same. This is one of the first examples of analysis I've seen that gets right to it. It's posing the question of if the "next" marginal unit of AI "power" - between 1 bit and 1.58bit - is worth the marginal energy-required delta. This is the kind of discussion we need to be having. Given the #externalities in play, regulation is, unfortunately, necessary. Instead of coming to this reality the "hard way", by finding ourselves with the aftermath of an unwieldly, chaotic energy/cpu/gpu rush which would surely be very inefficient (energy and otherwise)....maybe we can gaze to the horizon a bit here, and set things up to be more thoughtful/intentional before it's harder to course correct? We'd all love to have infinite energy. We don't. AI is fun and exciting, but we have to be grown ups and realists about how much "room" we have to just...go nuts with it. Steering toward 1-bit LLMs is the kind of rational choice we should take a very hard look at, even if it would be cooler to fire up those gpus/cpus to 11.

The day we've all been waiting for has arrived. A venture-backed BOS (non-)racking company has formed a JV with a first nation to manufacture a product neither party has any experience manufacturing (#polysilicon #pvmodules), while relying on untested #electrolyzer/ #hydrogen technology, using human labor where robots work better, in a country that - lets be honest - doesn't have much business manufacturing a largely commoditized product at this stage of its economic life. And, moreover, in an industry/sector that struggles to pull off even basic things, like getting it's most progressive and #solar-friendly state (until today, obviously) to implement an already-passed layup #communitysolar program based on wildly successful programs in other states (stay hot California Public Utilities Commission!). What a time to be alive! Naturally, this will probably come off without a hitch. And I will be genuinely thrilled if it does. We gotta try some different stuff in this here #energytransition, as the current trajectory, though a bit better than pre #IRA, still lands our kids in a hot mess. So, screw it ... go get em' Navajo Nation and Erthos!

Feels great to be back together with my man Lambros Theofanidis! Lambros was my first hire at Cypress Creek Renewables, which to this date remains one of my better life choices. He's a jack of all trades, and master of many...in addition to having a deft touch with partners, counterparties, and other key stakeholders. Plus he's just fun to work with. Welcome Lambros!

This piece reminds me, once again, of the importance of quality longform journalism. Julian Spector frames the opposing "takes" on the #ercot #bess #energystorage (bulls vs. bears) very effectively here, with insightful support from the proverbial "smartest guy in the room" on the topic (Zubaty). I land on the bull side. The fallacy of confusing "storage projects in the queue" for "future capacity" is at the heart of this view. Few would disagree that the fundamentals of i) load growth, and ii) increasingly intermittent energy mix offer a very compelling backdrop for continued excellent BESS returns. The countervailing force is, of course, the supply of new storage installations expected in the years to come. There will be a lot, that is certain. But if history is any indication, it will be less than queue-based predictions would have you believe. You know what few do? Compare "actuals" to previously projected capacity buildouts. As the article notes, BloombergNEF estimates that the 2024 tally will be 4.3 GW of BESS capacity additions. EIA guesses 6.4 GW. I went back 5 years and looked at EIA's 1-year-prior and 2-years-prior projections for ERCOT BESS capacity additions and compared that to what actually happened. 10 out of 10 of those projections were higher than the actual. Some were pretty close (for example their 2022 prediction for 2023 additions was only ~5% high), but a few were off by 50%+. (I haven't sleuthed around for the same BNEF comparison yet). This matches the "felt" experience as a market participant. Rumors of the imminent saturation of the ERCOT storage market are usually greatly exaggerated. I believe this is because they rest on a false premise: faster/more capacity buildout than what comes to pass. As I've mentioned before, I believe this arises from a fundamental lack of appreciation for how hard it is to shepherd a project to placed-in-service. The low barriers of entry in ERCOT are conflated with ease of development/EPC/financing execution. Not to mention a failure to appreciate what it takes to pull off the successful operation of an ERCOT BESS asset - specifically optimizing the revenue line. It's hard. Yes, even with your "proprietary algorithm". This systemic analytical "trap" is actually rather welcome for those of us continuing to develop and build ERCOT storage - it keeps the opportunity window open. But in the long run, I'd rather the confusion around this dissipate, so that the market find some footing beyond the always-over-correcting-one-way-or-the-other state it's in now. https://lnkd.in/eKaHeqYu