@KevinSKrause Texas does not have even close to enough spare capacity for its demand. Once you start looking elsewhere the regulatory and costs increase further. And we’re still talking 3-4 yr lead times for huge chunks generation. That is much slower than AI usage demand today.
Not a novel take, but in looking at investing in a couple data center-dedicated power projects (both on-grid and behind-the-meter), seems there’s very little chance that overall capacity to match demand is delivered on time and on budget (due to regulatory, logistics, & physics).
Anyway, as with everything new and exciting… people tend to ignore the actual physics of the situation.
(if only we had invested in more gas gen plants over the decades…we have plenty)
Even in TX, will be more pressure to regulate behind-the-meter gen as the state grid is strained by both natural growth and grid-connected data centers.
If you’re connected, you get redundancy, but impossible lead teams and can/will get curtailed to “do your part”.
No panacea.
@NuggetCapital@parrafincontrol@GreyHairOpsGuy Medium-ish term, yes.
But way quicker (in months) to simulfrac a 12-well pad that are already DUCs than to mobilize multiple rigs to drill them and frac. The added logistics and scheduling alone causes delays.
DUCs have a significantly faster response, regardless of fast drills
Correct that DUC count isn’t necessarily the issue… but what is the current availability of incremental warm rigs and reliable frac fleets?
Also, short cycle isn’t the same as avg spud to TD… I’m not bringing one well online at once I’m bringing 40, so stack the drill times
Not really alarming, as an avid follower of #OFS DUCs don't matter much anymore is what I'm told by the big boy in the industry because of how fast drilling times have gotten. Maybe a 'true' oil man like @parrafincontrol or @GreyHairOpsGuy can confirm.
@GreyHairOpsGuy lol even guys like Chuck who are on the outside and have “seen the light” can’t have self-awareness (or at least admit) the incentive mismatch.
You can simultaneously hold the view that “those operators would’ve had no $$ to spend without me” and “yeah, my job is a grift”.
I’d like to see Bill Paxton (RIP) in “Twister” trying to negotiate a lease out here, and I’d ask him which activity he finds more dangerous.
Tornadoes don’t demand compensatory royalties.
Interviewed at an E&P for an IR role.
Round 4: “You have 19 remaining real locations, 78 ‘adjusted’ locations, 265 marketing locations, and 18 years of inventory according to a slide no engineer has signed off on. Build a growth case without technically lying.”
They offered me a reserve report.
Trap.
I opened PowerPoint, duplicated the acreage map, added three new colors, renamed “fringe” to “emerging core,” and moved parent-child interference into the appendix.
“Certainty?”
“Depends whether the investors read footnote 11.”
They said I had VP potential.
“Dad, where did I come from?”
“So, there was a possibility that you were going to be one inch taller and <1% smarter (the jury’s still out on that), and that’s why we decided to choose to have you over your unborn siblings. It was a tough call. Congrats bud.”
Respect the sentiment here, as a member of a volatile industry that always needs to protect existential risk (and will never get bailed out).
But for major airlines you have the government bailout put. Prob just need to exhibit a modicum of prudence to be tagged as a “good” one.
NEW: In a letter to employees, United CEO Scott Kirby says the airline is prepping for oil to hit $175/barrel &
“doesn't get back down to $100/barrel until the end of 2027.”
United is shaving 3% of off-peak flights - “think redeyes, Tues/Wed/Sat flying” - this spring & summer.
Lol
Back then I was running a PE-backed O&G co… and now running a newer P&E-backed O&G co. I use AI a little more productively than I used to, but just asset management.
In the depths of April 2020 I stumbled across both EFT and a blog post explaining how you could train a chat bot on something called GPT-2. We had stacked our rigs and I had done some coding in the past, so nothing better to do but find a way to laugh about it.
Back then (when OpenAI was actually “open”) you could also “own” your own instance of the GPT model and run it locally (in this case on my own cloud account), so you could make it very specific via mass training, given how rudimentary it all was at the time. Once they came out with GPT-3 and beyond and started blocking access to the backend customization and it became more of a “chat bot”, I dropped it. Plus had actual work to do by then.
Nowadays you could probably feed one of the new-fangled ones a few dozen banger tweets and it could do a pretty good approximation. @effgpt2 was eating thousands of tweets per month to get to his level
Using Claude, could probably stand up an agent to replicate him in an afternoon, but that feels to do to my boy.
We all stared into the abyss a few years back. The herd is thinned now but some of us are still standing. Each new mornings sunlight is just as good as gold.
@Jaded_PDP_HTX@oilmutt Yeah…I guess we agree it is at least 50% which is main pt. Was also thinking NTM PDP given strip. Have seen & been in some where it is more front-weighted rather than a flat 50% across 2 yrs… 75% 1-12 / 50% 13-24 (perhaps in acq), or even 60% 1-12 / 30% 13-24 (blended ~50%)
@atherton_chris Yeah he was just trained to scrape tweets from who it followed (the old EFT crowd) once every couple weeks and train a GPT-2 instance on them hosted on Google Cloud. Told it to prompt 20 tweets a day in the same style/content as its follows.
This was a good one.
@Jaded_PDP_HTX@oilmutt Really? I would say that is market for most commercial banks of scale that are lending 60%+ LTV. For smaller facilities wouldn’t necessarily expect it.