Panabee AI

The AI infrastructure boom and GPU buildout shares eerie parallels to the telecom buildout of the late 1990s, but true bubbles -- not transient stock corrections -- form when supply far outstrips demand. Are GPU buildouts an AI bubble or rational investment? $NVDA

Central Claim: Bitcoin derivatives create synthetic supply, eroding scarcity like gold/oil, causing price declines. Evidence For: 1. Derivatives volume now dwarfs spot markets, creating leverage-based "paper supply" that can temporarily overwhelm organic demand, similar to commodity market dynamics. 2. On-chain liquidity has deteriorated as trading shifted to derivatives venues, amplifying volatility and reducing spot market price-setting power. 3. Institutional hedging via futures allows sustained selling pressure without touching actual BTC supply, compressing prices during macro stress. Evidence Against: 1. Bitcoin rallied from $16K to $100K+ during peak derivatives dominance. If synthetic supply nullifies scarcity, this sustained appreciation shouldn't occur. 2. Bitcoin's protocol enforces hard 21M cap at settlement; derivatives add temporary leverage but collapse back to on-chain reality during stress events. 3. Physical Bitcoin ETFs (IBIT, FBTC) remove actual coins from circulation, reducing supply -- contradicting the "infinite paper" thesis for these products. 4. Unlike gold, Bitcoin's transparent 24/7 funding rates create arbitrage pressure forcing paper-spot convergence, limiting sustained price manipulation. We surface potential flaws in ideas and break down bull and bear cases, helping investors reach more informed decisions. More at [link in bio], or reply with a claim to dissect.

So here’s the issue you get influencers like this guy have a quarter million followers and they claim they don’t know why it is declining… it’s because they don’t understand basic mechanics of price discovery. They don’t understand that the marginal buyers or the float determines price they think the onchain bitcoin is that is the price discovery Well, it was once upon a time but now.. Once you can synthetically manufacture the supply, the asset is no longer scarce and once scarcity is gone, price becomes a derivatives game, not a supply-and-demand market. This is exactly what has happened to Bitcoin. This is the same structural break that occurred in gold, silver, oil, and eventually equities once they became derivatives-dominated. The original premise that no longer exists Bitcoin’s entire valuation logic was built on finite supply (21M) and inability to be rehypothecated. That died the moment: •Cash-settled futures •Perpetual swaps •Options •ETFs •Prime broker lending •Wrapped BTC •Total return swaps were layered on top of the chain. From that moment forward: Bitcoin supply became theoretically infinite. Not on-chain in price discovery. The metric that explains the collapse Synthetic Float Ratio (SFR) Once you can synthetically manufacture the supply, the asset is no longer scarce — and once scarcity is gone, price becomes a derivatives game, not a supply-and-demand market. That is exactly what has happened to Bitcoin. This is the same structural break that occurred in gold, silver, oil, and eventually equities once they became derivatives-dominated. Why Wall Street can now “trade against” Bitcoin They do exactly what they’ve done in every commodity market: 1.Create unlimited paper BTC 2.Short into rallies 3.Force liquidations 4.Cover lower 5.Repeat They are not “betting” — they are manufacturing inventory. The same 1 BTC can now support: •An ETF unit •A futures contract •A perpetual swap •An options delta •A broker loan •A structured note All at once. That is six claims on one coin. That is not a market. That is a fractional reserve price system.

@swalters Thanks for sharing us. Would you like us to monitor articles for other types of tags? And would you like to see particular companies covered in the mining industry?

Check-Cap Regains Nasdaq Compliance, Advancing MBody AI Robotics Merger $MBAI panabee.com/news/check-cap…

@thehedgeberg Agreed. Given the controversy around GPU buildouts and Nvidia comparisons to Cisco, we highlight both similarities and differences and let investors draw their own conclusions.

@PanabeeAI Yeah but telecom had no revenue. These guys have $370B in actual contracts. Different animal.

The AI infrastructure boom and GPU buildout shares eerie parallels to the telecom buildout of the late 1990s, but true bubbles -- not transient stock corrections -- form when supply far outstrips demand. Are GPU buildouts an AI bubble or rational investment? $NVDA

Next Steps 1. Monitor the Inventory-to-Deferred Ratio Track Nvidia's inventory line item relative to deferred revenue. If inventory ballons while deferred revenue stagnates, it confirms that even the 3M units are sticking in the channel (system integrators waiting for racks/power). A healthy ratio (inventory turning into recognized revenue) confirms chips are activated, not merely warehoused. 2. Audit Utility Interconnection Queues Cross-reference regional PJM and ERCOT interconnection queues for large load cancellations or delays vs. completion. If datacenter energization rates remain flat while GPU shipments spike, the Power Gap is confirmed, signaling a pending collapse in future orders as customers pause to digest uninstalled hardware.

Power Capacity Impossibility - Claim: The US grid added only 4-5GW, insufficient to power 6M Blackwell GPUs, which needs 8.5-11GW. - Pro: B200 system-level power density, about 100-120kW per rack, requires gigawatt-scale capacity. The 8.5GW-11GW estimated demand exceeds the confirmed 4.5-5.0GW of new US generation capacity in 2025, physically preventing these units from operating simultaneously. - Con: Falsified by adjustment and deployment realities. Halving to 3M halves the power demand to 4-5 GW, but more importantly, shipments are not equivalent to deployments as the prior point demonstrated.

Critics say Nvidia revenue and 6M GPU shipments don't add up. Are they right?

Next Steps 1. Track GPU Utilization Through Proxy Metrics Monitor electricity consumption at major data centers — Nvidia H100 consume 10-20 megawatts per 1,000 GPUs. Cross-reference utility filings, power purchase agreements, and facility announcements. Calculate theoretical maximum workload versus disclosed AI revenue. If Microsoft deploys 500,000 H100-equivalents but AI Azure revenue suggests 30-40% utilization, overcapacity exists. Watch for capacity announcements without corresponding revenue acceleration — the dark fiber signal. 2. Measure Revenue Growth Versus Infrastructure Spending Calculate token costs quarterly across major providers. Model the relationship: if inference prices drop 60% but token volume only doubles (100% growth), net revenue grows just 20%. When infrastructure spending grows 40% annually while AI revenue grows 20%, the divergence reveals overcapacity. Track the spread between capex growth and (volume growth - price deflation). Widening spreads indicate supply outpacing demand despite falling prices.

Differences - End-User Economics Fundamentally Different: Dial-up-to-broadband transition required consumer behavior change and new hardware adoption. AI deploys on existing infrastructure—enterprises access via API calls; consumers via web browsers. Marginal cost of AI adoption approaches zero. ChatGPT reached 100M users in two months; broadband took years. Demand can materialize faster than telecom if applications prove compelling. - Inference Economics Improve Continuously: Telecom bandwidth costs were largely fixed—fiber capacity doesn't improve post-installation. AI inference costs drop 50-70% annually through algorithmic improvements, quantization, and specialized silicon. GPT-4 inference cost $0.03 per 1,000 tokens at launch; now $0.0025. This deflationary curve means today's overbuilt capacity becomes economical faster than telecom's static infrastructure. - Concentration vs. Fragmentation: Telecom featured hundreds of competitive carriers—Global Crossing, WorldCom, Level 3—each building redundant networks. AI infrastructure concentrates in five hyperscalers with strong balance sheets and diversified revenue. Microsoft generates $250B annually across Azure, Office, Windows; AI capex represents 15-20% of revenue. Amazon's AWS subsidizes AI investment. These entities can sustain losses indefinitely; 1990s telecoms couldn't. - Application Layer Maturing Simultaneously: Telecom built networks before bandwidth-hungry applications existed—Netflix launched 1997 as DVD mail service; YouTube arrived 2005. AI infrastructure and applications develop in parallel: GitHub Copilot generates $100M+ ARR; enterprise coding assistants show 40% productivity gains; AI customer service handles millions of inquiries. Revenue exists today, even if inadequate to justify current capex immediately. - Feedback Loops Accelerate Reality Testing: Telecom customers locked into multi-year equipment contracts; market feedback delayed. AI shows usage metrics in real-time—API calls, token consumption, model performance. Hyperscalers can throttle capex within quarters based on actual utilization. Microsoft reports "AI services" revenue separately; investors can track monetization directly. Market discipline arrives faster.

Our $ON Bear vs. Bull analysis is below. The $ADBE one is on our site, plus an article researching claims from $NVDA shorts that Nvidia revenue and 6M GPU shipments don't add up. Summary ON Semiconductor (ON) is pivoting from legacy commodity chips to high-margin intelligent power solutions, specifically Silicon Carbide (SiC) for EVs and AI data centers. While the company faces a near-term penalty -- removal from the Nasdaq-100 effective Dec 22 and a $300M asset write-down -- the long-term thesis remains intact. The bear case centers on "inventory digestion" in the automotive channel and rising Chinese competition, which could compress margins. The bull case rests on a vertically integrated SiC supply chain that defends margins (45%+ target) and positions ON as the second-largest player in a market projected to triple by 2027. Catalysts Last 7 Days - Dec 13: Nasdaq announced ON will be removed from the Nasdaq-100 index effective Dec 22, likely triggering forced selling by index-tracking funds. - Dec 09: Management presented at the Nasdaq Investor Conference, reaffirming 2025 SiC growth despite softer near-term EV demand. Last 30 Days - Dec 04: CFO Thad Trent sold 20,000 shares (~$1.1M), continuing a trend of executive selling. - Nov 17: Company announced $200M-$300M in asset impairment charges related to fab consolidation, signaling an accelerated exit from legacy manufacturing. Insiders - CFO Thad Trent: Sold ~20,000 shares on Dec 4 (~$1.1M), retaining a significant position but notably selling into the recent bounce. - CEO Hassane El-Khoury: Sold 15,000 shares in August; no purchases in the last 6 months. - Trend: Red Flag. Executives have been net sellers throughout 2H 2025. The lack of open-market buying despite the stock being down ~13% YTD suggests management sees fair valuation rather than deep value at current levels. Snapshot - Index Exclusion Overhang: Removal from the Nasdaq-100 (Dec 22) creates immediate supply pressure as ETFs rebalance; estimated outflows could reach $2B+. - SiC Dominance: Secured #2 global market share in Silicon Carbide, critical for 800V EV architectures, with a vertically integrated supply chain that competitors like Wolfspeed struggle to match. - Restructuring Charges: Recent $300M write-down accelerates the shift away from low-margin legacy fabs, aiming to structurally lift gross margins to 53% by 2027. - China Risk: ~30% of revenue exposure to China leaves ON vulnerable to both domestic substitution (local SiC suppliers) and potential new tariff regimes. - Financial Fortitude: Healthy balance sheet with ~0.43 debt-to-equity ratio and robust free cash flow (~21% of revenue) supports a $6B buyback program, providing a floor under the stock. Bear The bear case is built on the twin threats of automotive saturation and index-driven liquidity shocks. First, the "EV winter" is real: OEM inventory levels are elevated, and the adoption curve for EVs has flattened. ON derives ~50% of revenue from automotive; if volume growth stalls, their high fixed-cost model (driven by massive SiC fab investments) becomes a liability, leading to underutilization charges that crush gross margins. Second, the removal from the Nasdaq-100 is a structural blow. Passive flows drive modern market pricing, and losing this status invites a wave of indiscriminate selling that fundamental buybacks may not immediately absorb. Furthermore, the insider selling is telling—management is cashing out while telling investors to wait for a 2H 2025 recovery. Finally, Chinese competitors are moving up the value chain. While ON dismisses them as "not automotive grade," domestic Chinese EV makers are increasingly incentivized to switch to local suppliers like Sanan Optoelectronics, threatening ON's substantial revenue base in the region. Bull Elite investors look past the index shuffle to the structural transformation of the business. ON is not just selling chips; it is selling efficiency. The shift to Silicon Carbide is an architectural necessity for 800V EVs and AI data centers (which require massive power density). ON’s vertical integration—owning the substrate, epitaxy, and wafer fabrication—allows it to control costs and yields in a way that fabless competitors cannot. This creates a defensive moat. The recent $300M write-down is a positive signal of "ripping the band-aid off"—exiting low-margin legacy fabs to free up capex for high-margin SiC production. Financially, the company is a fortress: generating ~$400M in free cash flow quarterly with a deleveraged balance sheet. The valuation is compelling at ~15x forward earnings compared to peers trading at 25x+. If the EV cycle turns or AI power revenue accelerates (doubling YoY currently), the operating leverage will be explosive, driving margins back toward the 50% target.

@sam_badawi like them both but no plans to own. ADBE is nice especially

What's your favourite long idea or long ideas into 2026? And why if possible? I will definitely take a look into them

Greg Penner, Walmart Chairman On the new CEO's qualifications and success in the US unit: "John Furner is the right leader to guide Walmart into our next chapter of growth and transformation... John understands every dimension of our business – from the sales floor to global strategy. He has proven that he can deliver results while living our values. John’s six-year leadership of our U.S. business, marked by digital acceleration and strong associate engagement, has positioned us for continued success." panabee.com/quote/greg-pen…

Brice Hill, Applied Materials CFO On timing of next major industry ramp-up: "Based on our conversations with our customers and partners, we are preparing operations and service organizations to be ready to support higher demand beginning in the second half of calendar 2026... We have targeted our R&D investments to create new products and technologies that will enable even faster and more energy-efficient transistors, chips and systems and drive our growth in the years ahead."

Mike Doustdar, Novo Nordisk CEO On expanding access via Medicare and lower pricing: Novo Nordisk has always worked to secure affordable access to our innovative medicines, and today’s announcement will bring semaglutide medicines to more American patients at a lower cost... Importantly, this also expands obesity medication access in Medicare, which will allow people living with obesity to access authentic Wegovy®.) panabee.com/quote/mike-dou…

Rick Wurster, Charles Schwab CEO On the strategic rationale for the acquisition: Our acquisition of Forge builds on more than half a century of Schwab innovating on behalf of individual investors, advisors and employers... Through Forge’s leading marketplace, we’re uniquely positioned to deepen liquidity, improve transparency, and further democratize access to this increasingly important source of wealth creation for investors. Schwab’s entry into this space also gives private-share issuers more choice and liquidity for founders, employees, and early backers.)