• 10 Posts
  • 838 Comments
Joined 2 years ago
cake
Cake day: October 23rd, 2024

help-circle

  • Monarchies over the ages are a bad example, because there was a sustainability mandate to pass off a good country as inheritence to heirs. A candidate for worse countries would be Idi Amin, and US allies, but for US allies, their alliance is forced due to CIA coups/colour revolutions, and the scale of Idi Amin’s pillage is a million x lower. The utter oligarchist/corporatist supremacist nazi economics and geopolitics of US, together with the pure Israel/Zionazi capture of every elected official is a corruption that is beyond any other in both scale, and in fundamentally voluntary complicit bribery scheme for pure bribery value, excluding Mossad operations such as Epstein. The late stage empire pillaging stage is next level. That Congo arrests officials for corruption proves low levels of extended corruption. ie. punishments happen.




  • The need for more GPUs/datacenters was based on xAI using them for their own models, instead of competing with new datacenters/GPUs in the rental market. GPU rental prices went down to below cost levels for H series after Anthropic deal. It has bounced back this last week, but rates are still in the “very well supplied” range. On demand rates half of Anthropic/Google rental deals. GPU rental rates are the best measure of AI bubble point. Concerted fraud is still possible to provide the illusion of “this is fine”. The google deal only starts in September, while Anthropic got some free GPU time when deal was announced.

    In a way, pumping more money into AI (through IPO) delays the bubble pop, even if NVidia doesn’t sell as many GPUs if their customers add them to rental market. Bubble can pop from 3 directions. Less GPUs/TPUs sold than forecast; datacenter overcapacity ; understanding poor economics/tokenomics of high competition LLM labs.

    The sad part is that the more money invested, the greater the funding for the fraud, and longer lasting fraud.


  • There are signs that IPO is too big for market. 30% retail, and $135/share fixed price are because demand is not present. The illegal private/SPV market is trading below $129, where SPV entity that was not a fraud about to go bankrupt, and confident in the IPO would buy back shares up to close to $135.

    This BS by google is desperation to pump their existing stake. Anthropic’s headline $1.25B/month deal is also a fraud. Big discounts until after the IPO, and same cancellation rights.

    xAI got a made up (no one putting cash in) $250B valuation in its merger. Including Twitter it has barely any revenue, and without having to rent gpu datacenter space, was still losing $10b/year for a not particularly competitive Grok model.

    What’s withholding SpaceX from selling more stock with an even shorter time until they can be dumped?

    97% of the stock is already issued. 3% goes to new IPO bagholders, shortly later, passive index forced buying from the 3% market. The 97% are the ones with accelerated dump rights, that kick in a percentage shortly after fraudulent accelerated index inclusion date.











  • Very difficult project to track data for. Major difficulty is that capex budgets are like share buyback budgets. Implementation speed subject to change. It’s hard to measure revenue/profitability because AI credits tend to be bonus perks for enterprise customers, and in Meta’s case they use it for better brainwashing/ad revenue.

    The best way to measure Bubbliciousness is trends in spot GPU rentals. Nvidia has been selling more datacenter GPUs than datacenters completion for over 1 year. Those GPUs filter down to tier 2 providers because they cost $1.30 to $2.70 per hour to just sit in a warehouse. Anthropic’s deal to replace grok at Colossus 1 has significantly reduced tier 2 provider usage, and those spot rates have collapsed below their runcosts. https://lemmy.ca/comment/23443450



  • More precise pricing trends from premium tier 2 networks, show demand has drastically fallen over the quarter. H200 very close to its bare runcosts. Theory is that Anthropic’s overpayment for Collosus 1 (xAI) capacity has drastically reduced utilization at cloud rental service.

    NVIDIA B200 Blackwell (192GB HBM3e)

    • Tier-1 Spot Mechanic: AWS introduced B200 spot availability (p6 family baseline). However, because Tier-1 spot pools are subject to extreme automated reclamation, they command a rigid premium.
    • The Rebound: B200 spot experienced a localized surge around March 13th due to massive API volume peaks. 
    Week Ending (2026) Tier-1 Spot (AWS/Azure) Tier-2 Spot (CoreWeave/Nebius) Tier-2 On-Demand (Nebius Menu) The True Market State
    Jan 2 $6.40 / hr $4.50 / hr $7.80 / hr Initial launch window; hardware access highly constrained.
    Jan 16 $6.40 / hr $4.20 / hr $7.50 / hr Data center pipelines face massive backlog queues.
    Jan 30 $6.40 / hr $4.00 / hr $7.50 / hr Tier-2 unallocated floor space begins opening up.
    Feb 13 $6.12 / hr $3.85 / hr $7.20 / hr Multi-agent frameworks consume near-term supply.
    Feb 27 $6.12 / hr $3.50 / hr $6.80 / hr Influx of new Blackwell nodes flattens spot markup.
    Mar 13 $6.12 / hr $4.10 / hr $6.50 / hr The Agentic Peak: Spot surges via automated bidding.
    Mar 27 $5.90 / hr $2.95 / hr $6.00 / hr High supply volumes trigger a localized margin correction.
    Apr 10 $5.90 / hr $2.40 / hr $5.50 / hr Nebius and Lambda drop public on-demand baseline rates.
    Apr 24 $5.56 / hr $2.06 / hr $5.50 / hr The System Bottom: Spot drops to its absolute low.
    May 8 $5.56 / hr $2.40 / hr $5.50 / hr Short-term enterprise fine-tuning contracts absorb space.
    May 22 (Current) $5.56 / hr $2.90 / hr $5.50 / hr Current Rebound: Spot firms ahead of June 1 price hikes.

    NVIDIA H200 Hopper (141GB HBM3e)

    • The Clearance Reality: Unlike the B200, the H200 has failed to rebound. Because its FP8 processing layout lacks Blackwell’s native execution upgrades, developers are abandoning the H200 for production serving.
    • Below Cost: Tier-2 spot has flatlined at $1.45/hr, forcing providers to eat losses relative to raw facility overhead just to prevent expensive liquid-cooled spaces from sitting entirely dark.
    Week Ending (2026) Tier-1 Spot (AWS p5e baseline) Tier-2 Spot (CoreWeave/Nebius) Tier-2 On-Demand (Lambda/Nebius Menu) The True Market State
    Jan 2 $4.20 / hr $2.50 / hr $4.40 / hr Highly stable; utilized as the core long-context architecture.
    Jan 16 $4.20 / hr $2.30 / hr $4.25 / hr AWS implements dynamic Capacity Block adjustments.
    Jan 30 $4.20 / hr $2.10 / hr $4.00 / hr Early enterprise teams migration toward Blackwell blocks.
    Feb 13 $3.90 / hr $1.95 / hr $3.95 / hr Minor spot stabilization as alternative backends fill up.
    Feb 27 $3.90 / hr $1.80 / hr $3.80 / hr Shift to newer precision matrices devalues older stock.
    Mar 13 $3.90 / hr $1.95 / hr $3.80 / hr Minor agent-driven peak provides short-term support.
    Mar 27 $3.85 / hr $1.70 / hr $3.65 / hr Massive bulk capacity deployments flood European hubs.
    Apr 10 $3.85 / hr $1.55 / hr $3.50 / hr Market signals show severe oversupply on legacy nodes.
    Apr 24 $3.83 / hr $1.45 / hr $3.50 / hr The Floor: Prices slide below break-even run costs.
    May 8 $3.83 / hr $1.45 / hr $3.50 / hr Capacity remains completely unallocated across major nodes.
    May 22 (Current) $3.83 / hr $1.45 / hr $3.50 / hr Current Stagnation: Zero rebound; structural value tier.

  • A weird point that Nvidia CFO made to say “Nvidia is awesome” is a claim that GPU rental rates are up year to date. There was a crash at end of 2025. The low for the quarter was Jan 1st. The high was March 10th at peak of openclaw frenzy (validated by openrouter charts). Current rates are lower than that peak. But also comparison to 2025 Q1 (what I thought CFO meant, rates are down significantly) For single GPUs.

    1. NVIDIA A100 (Ampere — 80GB SXM)

    • Q1 2025 Baseline: High: $2.40 | Low: $1.60 | Close: $1.85
    • Q1 2026 Window: High: $1.65 | Low: $0.80 | Close: $1.15
    • Current Normalized Rate: ~$1.07 / hr (Stable floor; primary use shifts to entry-level fine-tuning and quantized serving). 

    2.[NVIDIA H100 (Hopper — 80GB SXM)

    • Q1 2025 Baseline: High: $7.00 | Low: $5.50 | Close: $5.80 (Supply constraints started easing, down from the absolute peak $10/hr overcharges of late 2024).
    • Q1 2026 Window: High: $3.45 | Low: $1.70 | Close: $2.35 (Hit an absolute low floor of $1.70 in late 2025 before a 38% contract rebound in March due to an influx of video-generation workloads).
    • Current Normalized Rate: ~$2.49 / hr (The standard baseline workhorse for mainstream API serving). 

    3. [NVIDIA H200 (Hopper — 141GB HBM3e)

    • Q1 2025 Baseline: High: $5.20 | Low: $4.50 | Close: $4.80 (Extremely scarce; reserved exclusively for elite labs running early frontier training).
    • Q1 2026 Window: High: $4.40 | Low: $3.50 | Close: $3.80 (Inventory stabilized as neoclouds widely deployed HGX baseboards).
    • Current Normalized Rate: ~$3.39 / hr (The most cost-effective tier for high-concurrency FP8 deployment). 

    4. [NVIDIA B200 (Blackwell — 192GB HBM3e)

    • Q1 2025 Baseline: N/A (Sampling/Testing phase; unreleased to the public marketplace).
    • Q1 2026 Window: High: $6.11 | Low: $3.05 | Close: $4.95 _(Initial public availability; premium pric

    5. NVIDIA B300 (Blackwell Ultra — 288GB HBM3e)

    • Q1 2025 Baseline: N/A (In architectural development; unavailable for rental).
    • Q1 2026 Window: High: $8.50 | Low: $5.50 | Close: $7.25 (Early access provisioning; highly volatile due to constrained data center site capacity).
    • Current Normalized Rate: ~$6.10 / hr (Neocloud standard rate; pricing reflects the premium for its 288GB memory pool). 

    for clusters, google AI mode simply can’t provide accurate info. Some providers have fixed premiums, others 0 premium. Many never change prices but mass email promotional discounts. For all I know, this entire analysis could have been a halucination meant to drive my narrative. I have not verified most data claims made as it would be too much work. I imagine most of the specific ones are accurate, and single GPU rental rates are the dominant market in the US, and that data should be solid, but FIIK.