$IREN's Most Likely Partner for SW1: OpenAI
For a site as large as SW1, the fastest path to monetization would be partner and use cashflow to continue to expand it’s full owned CSP when Vera Rubins come out in 2027 on SW2 and future sites/expansions that @FransBakker9812 is finding clues for such as (2). @litigious_dulce explains the insane economics Vera Rubins and how Blackwells economics will pale in comparison (3) so it’s important for IREN to time their self-bought chips for Vera Rubins. I will explain why IREN-OpenAI makes sense for both parties.
OpenAI:
The first key fact is that a critical part of DeepSeek’s breakthrough was their direct use of PTX bypassing some CUDA functions. The second piece of the puzzle is AMD’s hardware benchmarks always look great but their interface layer, ROCm, is far inferior to CUDA for full utilization of the hardware. Now if DeepSeek is able to write custom PTX, OpenAI can modify/customize ROCm to create an optimized version for their inference. For OpenAI, this will save them a fortune over Nvidia GPUs for inference where horizontal scaling makes sense, and it’s evident why OpenAI announced to buy 6GW of Instinct GPUs starting with 1GW in H2 2026.
No power constrained neocloud $CRWV, $NBIS could take the risk to buy that many AMD GPUs and tie up all their power. $NBIS doesn’t even have close to 1GW much less 6GW. Heck even, AMD knows that OpenAI is basically the only customer in 2026 at this scale and is willing to give up to 10% of their company to AMD if the whole 6GW of their GPUs is eventually purchased. OpenAI will need a Neocloud to deploy the 1GW in H2 2026. Even if NBIS does managed some how manage to pull off 1GW from somewhere, signing it entirely away would mean giving up their PaaS aspirations. Although there are are candidates like 5C group, IREN is a front-runner due to SW1’s proximity to OpenAI’s $ORCL sites in Abilene. OpenAI’s job post show no hiring for IaaS operations and even if they did, they would in house the GB300s first. OpenAI is signing deals with Coreweave on higher margin GB300s and would in house the operations of GB300s before in housing the operations of Instinct GPUs. IREN is a perfect candidate for OpenAI to provide the GPUs and have IREN operate them.
IREN:
How would the economic look like in relation to other known deals? We know that the top-line ARR for NBIS/MSFT deal is 11.6m/MW-yr (17.4B/300MW/5yrs). However after accounting cost of GPU, switches, $CRDO cables, DataOne colocation, NBIS is looking at a profit estimate of ~4.832m/MW-yr which is still fantastic. IREN profit estimate is 7.232m/MW-yr of profit for its self owned GB300s (1). However to have the asset/cashflow leverage to quickly build out B200/B300/GB300 and eventually Vera Rubins for it’s Canadian Sites (160MW) + Childress (up to 750MW) + SW2 (600MW) while maximizing it’s fleet composition Vera Rubins in 2027, it makes sense to do a hybrid Colo/IaaS partnership with OpenAI. There’s not much figures for this, but a rough ball park would be 2.5m/MW-yr to 4m/MW-yr. For a 5 year deal, this would be 17.5B - 28B. Unlike $NBIS who is paying for GPUs and Colocation with DataOne, this would be mostly profit for IREN.
$ORCL, OPENAI & SOFTBANK JUST DROPPED A $400B BET ON AI INFRASTRUCTURE
Five new U.S. data center sites will take Project Stargate to ~7GW of planned capacity -- one of the largest utility buildouts in U.S. history.
Power is now the bottleneck.
$IREN: The Most Asymmetric Play on AI Infrastructure
A single 50MW lease proves they can deliver.
A Sweetwater 1 deal (1.4GW) before energization triggers full re-rating.
This chart shows a realistic path to $100/share as 2.75GW gets monetized.
📈 $20B+ upside
#openAi $ORCL
$IREN - great quarter, raking in cash as guided. Earned about $81m cash profits from mining -- far ahead of any other miner. All-in cash hashcost coming in at $22.86, which is just insanely good, and true to their guidance. Easily the best at pureplay mining right now.
Costs are clearly under control. Won't speak to HPC prospects, but those renderings look great.
Here are the numbers as I have them. Total cash hashcost ~$23, $/KWH numbers are great as well.
For comparison, here is $CLSK. Also a great operator, probably in 2nd place (haven't crunched all the miners' numbers yet). ~$33 hashcost -- commendable, but $IREN's lower J/TH fleet, and lower costs on a $/KWH, make $IREN the winner this Q.
And here's what a struggling miner looks like ($43 hashcost, high $/KWHs):
Note that $MARA's "direct" cost includes hosting fees, which constitute ~45% of their BTC mining. Actual figures would be lower, but they'd presumably see an increase in SG&A as they move from hosting to O&O. So just look at "total".
If you'd like to get an idea of how all the miners would stack up if they all had the same fleet, just look at $/KWH numbers.
---
Back to $IREN, I like this nice touch where they no longer apply an "AI deduction" to guided hashcost. Thank you!
Only soft point was they sold 10m shares in Q1. Not ideal conditions to be selling, given share price at the time.
Best in class mining.. now all eyes are on HPC.
$TMDX $50s ▶️ $120s. Trimmed 1/3 position today, but still holding most. Shorts are trapped and still ~25% of the float. 👀
Q2 on track for a beat after one of the best Q1 reports I've seen. $150+ incoming?
Added a bunch of $TMDX this morning around ~$80, I agree with @JonahLupton's thesis here. The more I read about this company and it's growth story, the more bullish I become. $200+ within 2 years? Seems like a layup.
https://t.co/rIvSV1mSyl
$TMDX up 8% today but this is just the beginning.
I think we’re going to see a lot of green days for $TMDX over the next 6-8 weeks before we get to Q1 earnings because I’ll be shocked if we don’t get a big beat & raise and investors want to be in the stock before that happens.
Not to mention all those $SWAV investors now need to find a place for that capital. Come join us on the $TMDX train to $200+ within the next couple years.
Frustrated that I roundtripped $TMDX, but the business has never been more solid. One bad quarter and -65%... I think it bottoms here.
Q4 should be amazing per flight data, this will be a great "January" effect stock IMO - it's now my largest position. I think $150 by EOY 2025.
$TMDX is the most impressive sub $5B stock in the market.
Rev $144M vs $123M est
EPS $0.70 vs $0.24 est
Net income $25.7M
Gross margins 67%
Net margins 18%
Guide up
It’s like a young $ISRG.
This is another stock with major short interest at 26%, we saw what $HIMS did at 30%.
$TMDX continues to have a very strong Q1 with flight volumes up 21% QoQ.
I also updated my spreadsheet today and we're at 26.0 flights per day (in March) with just 2.5 days left in the month/quarter.
This means $TMDX should do approximately 2,130 flights in Q1 which is 21% above Q4 flight volumes.
There are several other data points and variables we can't track but it certainly looks like $TMDX is headed for a monster Q1 earnings report, perhaps a 15-20% beat above current street estimates which now stand at $123.4M
Q4 revenues per flight was $69,220 and if you apply that to Q1 flight volumes it gives us $147.4M which would be a 19.5% beat above street estimates.
Unfortunately we can't track exactly what % of Q1 flights is internal (company owned planes) vs external (chartered planes) so my math above assumes that % stayed constant from Q4. Obviously if it changed in a meaningful way we won't know until they report earnings. If that % did change, I doubt it was by more than 5% in either direction.
If $TMDX does indeed report $140-150M in Q1, I would expect them to raise full year revenue guidance from 20-25% to 25-30% which means at the midpoint we're talking about $563M which means the stock is trading at 4.4x NTM EV/SALES which is simply too cheap in the medtech sector when they already have 20% EBITDA margins.
Personally I think $TMDX does at least 30-35% growth in CY2025 with EBITDA margins in the 20-22% range in which case $TMDX is trading at exactly 20x NTM EBITDA which is also too cheap for a company growing EBITDA at ~40% for the next few years as they launch OCS 2.0 and KidneyOCS plus increase efficiencies with their 21+ planes.
West Texas is quickly becoming the new data center capital of the world, and $IREN is growing with it.
We're set to deliver a 75MW liquid-cooled AI data center in the coming months, while the 2GW Sweetwater Data Center Hub, near the Stargate project in Abilene, is progressing well - with the first 1.4GW coming online in April next year.
https://t.co/rVjsKsijev
@OpenAI@sama@Oracle@larryellison@SoftBank