Tit For Tat

nothing posted is investment advice, just me documenting my research. do your own research.

Substack:

[Tit For Tat @GameTheoryPP]

A name change and a +114% difference 💪🏼 $bitf —> $keel .

[Tit For Tat @GameTheoryPP]

I’m long $BITF mainly because of the shift from pure BTC mining toward HPC/AI, with Moses Lake as the first real test case for GPUaaS. Management laid that out on the Nov 13 Q3 call when they described a roadmap to convert existing sites to high‑density AI workloads and prepare for future NVIDIA hardware cycles like Vera Rubin/GB300‑class GPUs.

At Moses Lake in Washington, they’re converting 18 MW of gross load (about 15 MW IT at 1.2 PUE) from mining into liquid‑cooled HPC/AI, targeting an industry‑leading PUE in the 1.2–1.3 range with completion around late 2026. What matters to me is that Bitfarms signed a fully funded, binding agreement for about 128 million dollars with a large US data‑center infra provider to supply all critical IT equipment and building materials for that 18 MW. That works out to roughly 7.1 million dollars per MW for materials alone, which I treat as the base before adding the remaining construction costs to get to the usual 10–11 million dollars per MW all‑in DC cost range for high‑density builds.

For the GPU economics, I’m basing my assumptions to the IREN–Microsoft GB300 deal. That contract is around 9.7 billion dollars over five years tied to roughly 200 MW of IT load, and disclosures around that deal imply about 76,000 GB300 GPUs, or roughly 380 GPUs per MW of IT. Using that 380 GPUs per MW number on 15 MW at Moses Lake gives me about 5,700 GPUs if the site is fully populated with GB300‑class hardware.

On pricing, I assume GPUaaS deals for liquid‑cooled GB300s land in a 10–12 million dollars per MW per year band based on where large AI cloud contracts are being discussed and the economics that IREN’s deal seems to support. If I take 10, 11, and 12 million dollars per MW per year and spread that across 380 GPUs per MW, it works out to roughly 3.0, 3.3, and 3.6 dollars per GPU‑hour at 100 percent utilization when I divide annual revenue per MW by 365×24×380. That’s the range I’m using to think about top‑line potential per GPU hour.

On the cost side, I’m using a per‑GPU all‑in cost of about 80,000 dollars for GB300s including ancillary equipment, which is consistent with what has been reported for large GB300 purchase programs, even though some very large blocks can be a bit cheaper. For financing, I assume a high single‑digit interest rate (around 9 percent) via leasing over five years, similar to how management at Bitfarms has talked about leasing GPUs rather than buying everything outright. For power, I assume roughly 1.2 PUE and about 3 cents per kWh in Moses Lake based on public commentary about their low‑cost hydropower and efficiency targets at that site.

For the DC itself, I model the 128 million dollar materials commitment at about 7.1 million dollars per MW, then add roughly 3 million dollars per MW of other construction and integration costs to land in the 10–11 million dollars per MW build cost range that is consistent with other high‑density AI builds. Since the 128 million dollar portion is fully funded through the supplier agreement, I don’t layer extra cost of capital on that part; I only model cost of capital on any remaining DC spend and the GPU leases.

When I put this together, I get a breakeven GPUaaS price per MW that’s around 10.5 million dollars per MW per year for a 5‑year term if Bitfarms finances essentially all of the GPUs and the DC is in that 10–11 million dollars per MW range. If they can secure a deal closer to 11 million dollars per MW per year, then after covering GPU lease payments, power, O&M, and remaining DC capex, there is meaningful margin left, and the full DC plus GPU stack is paid off over that first five‑year term. That’s the range where the model starts to look attractive to me.

[Tit For Tat @GameTheoryPP]

@CapstackCapital Thanks. I think anything under $10 is a good buy at the moment.

[Tit For Tat @GameTheoryPP]

Update after $TTI (Tetra Technologies) Q1 2026 earnings:

https://t.co/lX5yltSat0

[Tit For Tat @GameTheoryPP]

$NN Update:

NextNav’s latest OCUDU news reaffirms it's not only a GPS backup, it ties into the AI-native 5G/6G stack.

OCUDU (backed by DoD’s FutureG office and big RAN/AI players) is building the open, AI driven RAN software that carriers and defense will run. NextNav isn’t building that AI stack itself but it’s contributing its Positioning Reference Signal (PRS) tech so precise location and timing become native features of that stack.

That’s the nuance: OCUDU is the AI infrastructure, but if PRS is wired into the default open-source CU/DU code, NN’s 3D PNT can scale with every deployment that adopts OCUDU, especially for “physical AI” use cases (autonomous systems, defense, critical infrastructure) where resilient PNT is mandatory.

Financially, Q1/Q4 ’25 into Q1 ’26 showed NN still burning cash but progressing on spectrum, regulatory, and network build-out. Q1 ’26 EPS came in at -$0.12 vs -$0.14 expected, revenue of $995,000 vs $900,000 expected.

[Tit For Tat @GameTheoryPP]

Vital Farms ($VITL) update: the thesis broke.
Q1 EBITDA: $5M (2.7% margin) vs. 16.9% a year ago. Full year guide cut to $0-$10M from $105-$115M. Exiting butter. Stock at $8.

The ERP disaster wasn’t a two quarter problem, it was a full year margin wipeout. At $8 you’re paying ~$170M for the operating business on ~$800M revenue with $113M cash on the balance sheet. Not zero, but not a buy until Q2 confirms Seymour is on track and shelf recovery is evident.

[Tit For Tat @GameTheoryPP]

Vital Farms ( $VITL) has been hit hard, but the market is treating a temporary operational screw up like a permanent problem. The core brand is intact, the balance sheet is clean, and the next two quarters should be the last of the pain before a longerterm recovery.

On my new substack acct. I break down what happened with the ERP rollout, how inflation really hits premium egg demand, the valuation range, and why I think today’s price portrays an unrealistically bearish future.

Read the full breakdown here: https://t.co/3qFMJ3yKam

[Tit For Tat @GameTheoryPP]

$POOL Thesis now on Substack. Link below
https://t.co/CiQkrbcToq

[IREN @IREN_Ltd]

Sweetwater 1 has been successfully energized – a key milestone in the development of the broader 2GW Sweetwater campus.

@danroberts0101, Co-Founder and Co-CEO of $IREN commented:

“Delivering Sweetwater 1 substation energization on schedule reflects our disciplined execution, the strength of our supply chain relationships and the efficiency of our vertically integrated development model. It is another example of our ability to design and construct large-scale infrastructure reliably and at speed to meet market demand.”

Learn more: https://t.co/Bo0tSv67jM

[Tit For Tat @GameTheoryPP]

Yesterday NextNav ( $NN) stock jumped 30%, with no major public news. Oppenheimer raised their price target to $25, likely to justify institutional accumulation. I think insiders anticipate NextNav on the FCC agenda next month, Oppenheimer’s call gives them a plausible rationale to buy now. This is COMPLETE speculation btw.

[Tit For Tat @GameTheoryPP]

Today at 1:30 p.m. CST, the FCC will be announcing its April agenda. I’m expecting NextNav ( $NN) to be on there, if not this month, then next month.

In my opinion, it’s just a matter of time. NextNav’s largest debt holder is Josh Lobel, CEO of the Milken Family Office, who also sits on President Trump’s National Intelligence Advisory Board and advises on these kinds of issues. That means the CEO of a firm holding most of NextNav’s debt also advises the President, who in turn appoints and oversees FCC leadership, including Commissioner Brendan Carr.

In addition, OIRA met with NextNav about a week ago, which is a standard step every NPRM must go through before it can be cleared for approval.

I wrote a probability analysis yesterday not taking any of the points made above into account to keep things as conservative as possible, see dropbox link: https://t.co/FPahp8uh9t

[Tit For Tat @GameTheoryPP]

Vital Farms ( $VITL) has been hit hard, but the market is treating a temporary operational screw up like a permanent problem. The core brand is intact, the balance sheet is clean, and the next two quarters should be the last of the pain before a longerterm recovery.

On my new substack acct. I break down what happened with the ERP rollout, how inflation really hits premium egg demand, the valuation range, and why I think today’s price portrays an unrealistically bearish future.

Read the full breakdown here: https://t.co/3qFMJ3yKam

[Tit For Tat @GameTheoryPP]

@FriendlyCapMgmt In my replies I said I misspoke. The meeting was about NN, not with them.

[Tit For Tat @GameTheoryPP]

Today at 1:30 p.m. CST, the FCC will be announcing its April agenda. I’m expecting NextNav ( $NN) to be on there, if not this month, then next month.

In my opinion, it’s just a matter of time. NextNav’s largest debt holder is Josh Lobel, CEO of the Milken Family Office, who also sits on President Trump’s National Intelligence Advisory Board and advises on these kinds of issues. That means the CEO of a firm holding most of NextNav’s debt also advises the President, who in turn appoints and oversees FCC leadership, including Commissioner Brendan Carr.

In addition, OIRA met with NextNav about a week ago, which is a standard step every NPRM must go through before it can be cleared for approval.

I wrote a probability analysis yesterday not taking any of the points made above into account to keep things as conservative as possible, see dropbox link: https://t.co/FPahp8uh9t

[Tit For Tat @GameTheoryPP]

https://t.co/6VGrYw1wCc

[Tit For Tat @GameTheoryPP]

@popartpopp @MGardnerDA1235 No, it can be announced any month. It’s just agreed upon by a good few analyst that it’s likely to happen in April or May.

[Tit For Tat @GameTheoryPP]

@MGardnerDA1235 Agreed

[Tit For Tat @GameTheoryPP]

@MGardnerDA1235 Should be on it for May then. Disappointing .

[Tit For Tat @GameTheoryPP]

Sorry that was poor phrasing OIRA had a meeting about NN, not directly with them.

OIRA review is a required step for a significant NPRM. They meet with both supporters and opponents and then decide whether the draft can move ahead. Just because they had an opposition meeting doesn’t mean anything besides that the plan is moving forward in my opinion l.

[Tit For Tat @GameTheoryPP]

$NN

[Tit For Tat @GameTheoryPP]

https://t.co/6VGrYw1wCc

[Tit For Tat @GameTheoryPP]

https://t.co/CV9JkHEmFj

[Tit For Tat @GameTheoryPP]

I post a lot about high-beta companies, but the other side of that trade matters just as much: the hedge.

If AI drives a deflationary shock, weaker demand, falling prices, and rising unemployment, markets tend to sell growth assets and push policymakers toward easier policy and lower rates.

That’s where a duration hedge like $TLT becomes interesting. A reset in long term interest rates can generate gains in Treasuries at the same time high beta stocks and BTC are taking a hit.

It’s been good to see people like @infraa_ digging into similar themes and laying out the case in more detail.

[Robert (infra 🏛️⌛️) @infraa_]

https://t.co/I57C4sRNGf