rama reddi (@rreddi) header
rama reddi (@rreddi) profile picture

Automation, Robotics, Systems Integration

Texas, USA
Joined February 2009
Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)Photo by rama reddi (@rreddi)
I believe life is infinite, and we are here to pay it forward.
3
32
9,913
If you have power and fiber... you are sold.
Anthropic and NVIDIA are now hunting for data centers 5 MW at a time. Anthropic has $901B of compute on order and is still asking for 5 to 10 MW in old buildings. > NVIDIA has proposed converting 130+ old phone buildings in up to 20 states > Verizon says a converted central office sold out in 24 hours > Crusoe will build 3.5 MW data centers on a factory line 5 MW is about 2,500 GPUs and $125M a year in rent. OpenAI's Ohio campus is 850 times that size and still years out, so anything with power connected today gets rented.
73
$NBIS Probably explains the bloodbath today. One timers + 2027 capex
$NBIS is selling off because the quarter was very messy The revenue beat is low quality. Q3 came in at $981M versus about $910M expected, but $325M of that was one-time revenue from asset-light partners, where Nebius enables its architecture on capacity someone else paid for. Ex that, Q3 was about $656M, up roughly 13% QoQ, and a miss. The 2026 guide slid. Revenue now lands at the low end of $3.0 to 3.4B because colocation capacity moved into 1H27, including the rest of the New Jersey build for Microsoft. That implies about $1.04B in Q4, barely above a Q3 headline that had the one-off in it. Then they dropped ARR. The $7B+ ARR target moved from December to February, and in the same letter management said ARR is no longer the right metric. Retiring a metric the quarter you miss it gets noticed, even with a reasonable explanation (it annualizes one month, and asset-light revenue makes it choppy). Backlog barely moved. RPO, contracted revenue not yet recognized, was $38.7B, up only 3% QoQ, despite $12B of TCV signed since the start of Q3. Most of that was likely signed in Q4 (my read, they don't split it), but the reported number didn't grow. But the real reason is capex. 2026 capex rose to about $27B, above the $20 to 25B range. 2027 capex is $40 to 45B, against roughly $29B analysts were modeling three weeks ago. That is $11 to 16B of spend the street didn't have. They say 75 to 85% of 2027 capex is covered by prepayments and asset-backed financing on investment-grade contracts. That leaves $6 to 11B to raise elsewhere, from a company that already did $5.75B of converts in August, about $1.0B through the ATM in Q3 and a $3.5B Citi facility in October.
1
158
$40B plus capex 2027. 2026 revenue slippage into 2027. I guess there will be less hate on IREN now. Based on peer revenue multiples, IREN should be trading at $35 compared to NBIS $140.
2
209
Never a dull moment in the new world order. 🙂
Werey don spoil show😭💔
62
CoreWeave $CRWV CFO: "We expect the contracted cash flow from our existing customer commitments to be sufficient to service and repay the debt we have outstanding today as well as any additional debt we expect to raise to deliver these contracts." "That includes both asset-level secured debt and unsecured debt sitting at the parent level, and that is before counting renewals." It has already paid down more than $4B of debt.
15
27
9
374
56,108
rama reddi retweeted
BIPARTISAN SENATE BILL COULD FAST-TRACK AI DATA CENTER EXPANSION AND POWER INFRASTRUCTURE, POTENTIALLY EASING MAJOR DEVELOPMENT BOTTLENECKS; VOTE POSSIBLE AS EARLY AS NEXT WEEK — THE VERGE A bipartisan Senate permitting bill could accelerate U.S. energy and AI data-center construction while lowering electricity costs, but environmental groups warn it would weaken oversight and communities’ ability to challenge projects. •Faster Approvals: The Bipartisan American Affordability and Jobs Act (BAAJA) would require decisions within 45 days of completed environmental reviews, exempt more projects from federal review and limit legal challenges to 150 days. Courts could no longer halt projects solely for violations of the National Environmental Policy Act (NEPA). •Data-Center Concerns: Easier permitting could accelerate gas pipelines, fossil-fuel generation and data centers while restricting public challenges involving environmental damage, water use and pollution. •Consumer Protections: Data centers would pay for new and existing transmission infrastructure, provide financial guarantees against abandoned projects and face expanded electricity-use disclosure requirements. •Cost Savings: Independent modeling projects annual household electricity savings of $50–$78 by 2040, compared with a White House estimate of up to $251. The bill could also nearly halve transmission construction timelines. •Legislative Outlook: The bill has bipartisan support and could advance next week. Supporters anticipate faster renewable and conventional energy development, while critics want stronger environmental protections. Noise and on-site pollution concerns remain unaddressed.
5
9
6
45
28,135
Securing long term contracts while doubling GPU hour rates on H100 chip that has already depreciated 100% on the books. In this compute constraint environment, GPU's are here to print money for as long as 10 years.
One of the most interesting comments on $IREN’s earnings call came from Chief Commercial Officer Kent Draper: “This is genuine long-term contracting for those H100s that has still resulted in a doubling in price". IREN has re-contracted those H100s at DOUBLE their previous pricing. Crucially, Draper clarified that the higher price was secured under a multi-year agreement. That detail deserves attention. A central question for AI infrastructure investors is how much earning power GPUs retain as newer generations arrive. Here, we have a concrete example of older-generation hardware securing several years of demand at substantially higher pricing than its previous contract. For those machines, that could improve lifetime returns while reducing exposure to future renewal pricing during the new contract term. It also challenges the assumption that a new GPU generation automatically destroys the commercial value of the previous one. Different workloads have different requirements, and customers can continue finding value in existing hardware. Some perspective matters: we don’t know the GPU count, absolute pricing or full contract terms. Doubling the rental price does not mean doubling profits, and one agreement cannot establish the economics of an entire fleet. Still, this is a compelling commercial data point.
1
110
rama reddi retweeted
This is f*cking insane Nvidia CEO Jensen Huang just did the AI bubble math live: "They don't have to be anywhere near a trillion dollars in revenues to be successful" In 5 minutes, he walks through exactly why the "OpenAI can't pay for it" fear misses most of the build-out 0:00 The bubble question everyone is asking 0:40 Break it down to first principles 1:22 $3-4 trillion a year in infrastructure by the end of the decade 1:59 Moore's law is over, twice the compute now costs twice the money 3:17 $5 billion per gigawatt then, $50 billion now 4:43 Why two labs don't carry the whole trade When AI stocks drop, most people blame the labs' revenue The split they land on: general compute replacement → neoclouds, sovereigns and startups → frontier labs → a few trillion of infrastructure Old way: one lab misses a number, sell everything New way: ask which part of the build-out actually depends on it Better than any $2,000 finance course I've seen
55
171
15
1,383
168,830
$IREN Equity: $5.116B Market cap: $7.538B IREN just added one cool bollion equity in just one quarter alone. A business running at 86% GM is trading at 1.473 book value.
3
305
rama reddi retweeted
The governor-elect of Texas, Gina Hinojosa, has explicitly said that all currently permitted datacenters in Texas will be allowed to finish construction. This essentially covers all of the 2027 and likely well into 2028. She is going to pass laws that ensure that all *newly* permitted datacenters going forward for delivery in 2028/2029 will pay taxes, commit to no increase in power prices for consumers, don’t impact water availability for people, be constructed with union labor and give “Texans a seat at the table.” Sounds reasonable. She doesn’t want to lose a lot of good blue collar jobs and property taxes for small towns in rural Texas. Datacenter commentary starts at 6:30 in this interview. Also attaching her official campaign position from her website in the next post. Sidenote: After watching a bunch of her speeches and interviews, I do think she is a political talent and probably has a future in national politics. youtu.be/D-T6H3b8ZeE?is=mEWu…
1,635
2,205
1,045
17,456
18,410,107
$IREN massive earnings report. 90% GPU capex prepaid. $40M per MW shoJanuary. $22M per MW fiver year contract. $4B ARR on track for january.
7
859
rama reddi retweeted
1/ Everyone take a deep breath. Texas data centers aren't going anywhere. The Texas governor has very limited authority to stop them. Calls for a pause or moratorium must be approved by TCEQ—the permitting authority of Abbott appointees. Here's how Texas permitting really works👇
72
407
131
1,990
443,133
Gina Hinojosa’s win in Texas raises a questions for $IREN, given how much of its development pipeline is located in the state. But not all of $IREN’s Texas projects carry the same political risk. There is a big difference between infrastructure that is already built or progressing through the approval process and entirely new projects that haven’t entered that process yet. Starting with Texas: Childress is already an operating site, supporting $IREN’s Microsoft and NVIDIA deployments. It is not a new project waiting for a permit. Sweetwater 1 (1.4GW) has already been energized, and construction of the first 300MW of DC capacity is underway, with delivery targeted for Q4 2027, which is way further along than a project that exists only on paper. Sweetwater 2 (600MW) carries a slightly higher risk since it is not yet energized, but it has already been conditionally included in ERCOT’s Batch Zero as Base Load, alongside Sweetwater. Conditional inclusion is not final approval, and verification is still ongoing, but Sweetwater 2 has already entered that process and received a conditional classification. What can the new governor actually change? Hinojosa campaigned for a moratorium on new data-center development until lawmakers establish stronger protections around electricity, water and local communities. She can influence state permitting policy, direct state agencies within her authority, appoint PUCT commissioners subject to Senate confirmation, and push for new legislation. That could make future development in Texas considerably more difficult. But she does NOT personally decide ERCOT’s technical classifications, and her election does not automatically cancel existing Batch Zero classifications. There is still risk for projects already in the process. The verification requirements and remaining approvals have to be completed, and a new administration could change the broader regulatory environment. But I would separate that risk from the much greater uncertainty facing projects that haven’t secured a place in the existing process. And $IREN is not just Texas. With infrastructure in Canada, a major development pipeline in Oklahoma, and projects in Spain and Australia, its entire pipeline is substantially less dependent on the situation in Texas. There are also reasons to think $IREN may be better positioned to navigate the changing political environment than some of its competitors. $IREN has spent years building relationships with the communities... local hiring, investment in community infrastructure, grants to local organizations, workforce training and a closed-loop cooling system that doesn’t rely on evaporating local water are all part of its approach. In Childress, it has created more than 200 permanent jobs, many filled by local residents, and has also stepped in to help the community during its recent water difficulties. That doesn’t mean there won’t be opposition, but these efforts matter. The backlash against data centers is often about communities feeling they carry the costs without sharing in the benefits. $IREN has made a real effort to address that, and I believe this could work in its favor as scrutiny increases. BTW Texas isn’t the only state with a Democratic governor developing DCs. Virginia and Maryland are good examples. Both are introducing stronger rules around energy, water and community protections, but neither is abandoning the industry. Virginia alone has 391 additional data-center projects in its development pipeline, according to recent industry data. The direction moving forward is clear: DCs will increasingly have to demonstrate that their projects benefit local communities which can only benefit the established players that have already invested in infrastructure, built local relationships and demonstrated a willingness to work with regulators.
14
20
1
203
20,634
rama reddi retweeted
Elon Musk says the AI boom will hit a WALL by the end of 2026. He says the whole industry is about to learn a hard lesson in hardware. Chip output is exploding, but electricity is nearly flat everywhere outside China. So the two lines cross this year. He says by the end of the year, we will be making more AI chips than we can actually turn on. Millions of them pile up with nowhere to plug in. In his words, the bottleneck for the next year is not chips, it is raw power. Whoever can turn on the most chips fastest wins the race. In the end, the AI race comes down to one thing: Raw electricity. — Elon Musk (.@elonmusk) on Dwarkesh Patel's (.@dwarkesh_sp) podcast
Elon Musk says billion dollar AI companies with money to spend can't build power plants fast enough to meet AI's electricity demand. Every major power project requires approvals and equipment, including: • Switchgear • Gas turbines • Power plant permits • Electrical transformers • Utility interconnection studies • High-voltage transmission lines • Specialized turbine blades and vanes If even one component is delayed, the entire project stalls... leaving AI chips without electricity. Musk says gas turbines are sold out through 2030, with specialized blades and vanes being the biggest bottleneck. Building more power plants isn't the quick solution AI companies thought it'd be. — Elon Musk (.@elonmusk) on Dwarkesh Patel's (.@dwarkesh_sp) podcast
631
1,239
270
9,652
6,147,332
A must watch! Biggest take aways... ...Ninety-five percent of customers run modified models. Open inference remains under-optimized. ...Sole bottleneck is execution amid shortages of GPUs, power, CPUs, NVMe, and talent. youtu.be/qJkcjURcAe8?is=ByW1…
170
🚨 AI INFRASTRUCTURE BOOM JUST GOT BIGGER. GS has raised its 2030 global data center capacity forecast from 217 GW to 273 GW a massive 26% upgrade. That's an estimated $8.5 TRILLION in infrastructure spending over five years. And here's the real story: AI demand isn't slowing down. The race for power, GPUs, and data center capacity is accelerating. The biggest AI investment cycle may still be ahead of us. 🧵👇
22
88
19
488
50,078
🚨 NOMURA: THE MARKET MAY BE MISREADING THE ENTIRE AI DATA CENTER BOOM. Big Tech is spending hundreds of billions on AI infrastructure. Free cash flow is under pressure. Investors are worried the AI boom isn't profitable. But Nomura argues something very different: AI infrastructure is already highly profitable. The cash flows simply haven't caught up with the investment cycle. Here's the breakdown 🧵👇 1/ The profitability numbers tell a different story. According to Nomura's November 2 report: • AWS revenue: $46B, +39% YoY • AWS operating margin: 38.7% • Google Cloud revenue: $30.7B, +103% YoY • Google Cloud operating margin: 36.2% • Azure growth: +48% These are not the numbers of an industry struggling to generate operating profits. 2/ The real issue is timing, not necessarily profitability. Big Tech pays for data centers, GPUs, networking and memory upfront. Revenue and cash generation arrive over subsequent years. That creates a classic free-cash-flow J-curve: CAPEX accelerates → FCF deteriorates → Infrastructure generates revenue → Investment growth moderates → FCF recovers. 3/ Nomura expects operating income to surge before free cash flow. Its illustrative forecasts suggest operating-income growth could peak around 2029, while FCF growth accelerates later. That distinction matters enormously for investors valuing AI infrastructure businesses. 4/ Why can't hyperscalers simply stop spending? Because AI increasingly threatens their existing businesses. Search, advertising, shopping, software and cloud services are being reshaped by AI agents. Companies that underinvest today risk surrendering customers, data and market share tomorrow. 5/ Demand may not be the primary constraint. Capacity is. Nomura cites enormous contracted cloud backlogs: • Google Cloud: $560.8B • AWS: ~$496B • Microsoft commercial RPO: $677B These figures have different reporting definitions, but collectively illustrate substantial contracted demand. 6/ Big Tech has financing alternatives to cutting CAPEX. Nomura identifies four paths: ① Accelerate investment ② Freeze CAPEX ③ Use joint ventures and operating leases ④ Monetize contracted future cash flows Its preferred expectation is continued investment supported by increasingly sophisticated financing structures. 7/ Oracle offers a real-world example. Nomura highlights Oracle's combination of long-term leases, project financing, customer prepayments and customer-provided hardware. The broader message: hyperscalers can restructure who finances infrastructure without necessarily reducing how much infrastructure gets built. 8/ This distinction is critical for semiconductors and memory. Negative hyperscaler FCF does not automatically imply collapsing demand for GPUs, DRAM, HBM or other data-center components. If spending continues through financing structures, semiconductor demand can remain resilient even while hyperscalers optimize their balance sheets. 9/ What could invalidate the thesis? A genuine compute oversupply. Watch: • Cloud revenue growth • GPU rental pricing • Contracted backlog growth • Data-center utilization • Memory supply commitments • Hyperscaler investment guidance These are more useful indicators than negative FCF alone. 10/ The investment takeaway: The key question is not simply: "Why is Big Tech burning so much cash?" It is: "Will the infrastructure being built today generate enough durable cash flow to justify its cost of capital?" Nomura is optimistic. But the answer ultimately depends on pricing, utilization, investment discipline and the durability of AI demand. The AI infrastructure cycle may be a cash-flow timing problem rather than a profitability crisis. $NVDA $MU $GOOGL $MSFT $AMZN $META $ORCL
116
525
63
2,411
335,739
IT IS TIME TO SELL $IREN 8.5 Billion Market Cap 7.6 Billion Cash on Balance sheet 9.7 Billion dollar deal with $MSFT 4 Billion end 2026 ARR 14 Billion projected end 2027 ARR 6+ GW of capacity secured. Only 1 billion debt needs to be paid back before 2031 2% avg interest on the debt 4.2 Billion Shareholder Equity 500 mw to be signed in 2027 at 2x compute prices
93
65
7
953
193,297
Insatiable Demand for compute.
Same chips as May. 33% more per hour. @BUZZHPC on-demand rates: H200: $3.99 → $5.29 H100: $2.99 → $3.99 Today’s H100 rate is what an H200 cost six months ago. Every step up on that chart lands on GPUs $HIVE already owns.
79
rama reddi retweeted
AI capex projected to rise sharply next year. AI stocks are currently out of favour, won't last for long.
62
99
19
894
80,041