Helping investors decode business news and stock events. Tag @PanabeeAI for analysis. For info purposes only. Not investment advice. See website for more.
Joined May 2024
- Tweets 1,396
- Following 8
- Followers 96
- Likes 32
144 Photos and videos
Pinned Tweet
13 Dec 2025
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
1
1
161
13 Dec 2025
Critics say Nvidia revenue and 6M GPU shipments don't add up. Are they right?
1
59
13 Dec 2025
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.
1
41
13 Dec 2025
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.
42
13 Dec 2025
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
1
1
161
13 Dec 2025
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.
1
38
13 Dec 2025
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.
36
14 Nov 2025
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-penne…
46
14 Nov 2025
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."
38
14 Nov 2025
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-doust…
85
14 Nov 2025
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.)
46
7 Oct 2025
Why Is POET Technologies (POET) Up? | Oct 07 2025
Summary
POET Technologies (POET) develops optical interposer technology for data centers and AI applications, recently securing a landmark $75M investment to fund growth.
The stock saw a significant volume increase on Oct 07 following the funding news. Heavy insider selling, including by the CEO, and substantial shareholder dilution over the past year present potential risks for investors.
Catalysts
Last 7 Days
Oct 7: Announced the closing of a $75M private placement with a single institutional investor, the largest in its history. The deal includes 13.6M common shares and warrants at a combined price of $5.50 per unit.
Oct 1-2: CEO Suresh Venkatesan and board member Glen Riley sold a combined ~3.3M shares.
Last 30 Days
Sep 30: Launched 1.6T optical receivers for AI networks in collaboration with Semtech.
Sep 30: Won "Most Innovative" award at the ECOC 2025 exhibition.
Sep 29: Announced a collaboration with Sivers Semiconductors on external light sources for the AI market.
Insiders
Suresh Venkatesan (Executive Chairman & CEO): Sold ~3.3M shares for ~$20.2M on Oct 1.
Glen Riley (Board Member): Sold 8,163 shares for ~$50.7k on Oct 2.
Insiders have only sold shares in the past three months.
Snapshot
Fortified Balance Sheet: The recent $75M cash infusion brings the company's pro-forma cash position to over $150M with no significant debt, providing substantial runway for acquisitions and R&D.
Persistent Cash Burn & Minimal Revenue: Despite the new funding, the company remains pre-revenue at scale. Q2 2025 results showed revenue of only $268k against a net loss of $17.3M. The company has a history of negative net margins and return on equity.
Significant Dilution: The latest financing adds over 13.6M shares and an equal number of warrants. Shareholders have already been diluted by 29.3% in the past year.
Insider Selling Pressure: The CEO's large stock sale immediately preceding the financing announcement may signal a lack of confidence or simple financial planning, but the optics are poor for new investors.
PanabeeGuard
Grade: C
🚩 Heavy insider selling by top executives, including the CEO, is a significant red flag.
🚩 Consistent and substantial shareholder dilution to fund operations.
🚩 Minimal revenue generation despite years of operations.
Newcomer Notes
The company operates in the high-growth AI and data center markets, but it is a development-stage company with significant financial losses.
The recent large investment provides capital for growth but also causes immediate and potential future dilution through warrants.
Pay close attention to the company's ability to convert its technological developments into meaningful revenue and the rate of its cash burn.
103
6 Oct 2025
Why Is OneMedNet (ONMD) Up? | Oct 06 2025
Summary
OneMedNet (ONMD) is a healthcare data company providing access to de-identified clinical and imaging data through its iRWD™ platform for the life sciences industry. The stock surged over 100% on Oct 06 following a partnership announcement with Palantir.
Beware the firm's significant financial distress, with declining revenue and substantial net losses raising questions about its operational viability.
Catalysts
Last 7 Days
Oct 6: OneMedNet announced a multi-year agreement to use Palantir's Artificial Intelligence Platform to enhance its healthcare data analytics offerings.
Last 30 Days
Sep 24: Announced a strategic partnership with Medcase to enter the $1.5B healthcare data annotation market.
Sep 15: The company regained compliance with Nasdaq's minimum market value listing requirement.
Insiders
No notable insider activity in the last 30 days.
Snapshot
Palantir Partnership: The primary driver is a new partnership integrating Palantir's AI platform with OneMedNet's network of 1,750 provider sites and 131M clinical exams. This deal aims to accelerate data discovery for life sciences clients.
Dire Financials: The company's financial health is precarious. Trailing twelve-month revenue was just $459k, a ~60% year-over-year decline, with a net margin of -728.7% and an operating margin of -2030.4%.
High Valuation: The stock trades at a Price-to-Sales ratio of ~74, near its two-year high. This valuation appears heavily dependent on the success of new partnerships rather than existing business performance.
Weak Cash Position: The company has a net cash position of -$278k ($122k in cash vs. $400k in debt) and a free cash flow of -$8.21M over the last twelve months, indicating a significant cash burn rate.
PanabeeGuard
Grade: D
🚩 Extremely high Price-to-Sales ratio suggests a valuation disconnected from current financial reality.
🚩 Severe revenue decline and substantial net losses point to fundamental business challenges.
🚩 Today's price may be momentum-driven based on the Palantir news, creating high volatility risk.
Newcomer Notes
The investment thesis hinges almost entirely on future potential from partnerships, particularly with Palantir, not on current business fundamentals.
The vast gap between the company's market valuation and its actual revenue and profitability presents a significant risk of a value trap.
1
70
6 Oct 2025
Is Eos Energy Enterprises (EOSE) a Buy? | Oct 06 2025
Summary
Eos Energy Enterprises (EOSE) designs and manufactures zinc-based battery energy storage systems for utility and industrial applications, an alternative to lithium-ion technology.
After a strategic partnership announcement, the stock saw a rise on Oct 06, but significant risks remain, including ongoing securities fraud investigations by multiple law firms and substantial insider selling.
Catalysts
Last 7 Days
Oct 06: Stock rises on news of a strategic partnership with Unico to enhance its energy storage solutions.
Oct 03: Filed a shelf registration to potentially offer $2.8M in common stock.
Last 30 Days
Sep 08: Launched DawnOS™, a new proprietary battery management and software platform.
Sep 05: Jefferies initiated coverage with a "Hold" rating and a $6.50 price target.
Sep 03: An insider sold shares worth $395,000.
Insiders
- Alexander Dimitrief, a director, sold 45,000 shares on Aug 05, 2025.
- Marian Walters, a director, sold 50,000 shares on Sep 03, 2025.
- Over the last three months, insiders have sold 659,074 shares and have not purchased any.
Snapshot
- Eos is experiencing rapid revenue growth, with Q2 2025 revenue of $15.2M nearly matching the full-year 2024 sales. The company reaffirmed its 2025 revenue guidance of $150M-$190M.
- Despite revenue growth, the company has a significant cash burn and widening gross losses. It holds a conditional commitment for a loan of up to $398.6M from the U.S. Department of Energy, which is still subject to conditions.
- Multiple law firms are investigating the company for potential securities fraud following the abrupt termination of its CFO in May 2025 and allegations of misrepresenting its sales backlog.
PanabeeGuard
Grade: D
🚩 Multiple law firms are actively investigating Eos for potential securities fraud and misleading statements.
🚩 A class-action lawsuit was filed alleging the company overstated its backlog and the security of purchase commitments.
🚩 Significant and consistent selling by insiders over the past year with no open market buys raises concerns.
🚩 The abrupt termination of the CFO in May 2025 triggered a stock drop and intensified scrutiny.
Newcomer Notes
- The energy storage sector is highly competitive, with Eos facing established lithium-ion technologies.
- The company's valuation appears high relative to current sales, embedding expectations of massive future growth.
- Ongoing legal investigations present a significant, unpredictable risk that could materially impact the stock.
1
61