|
Feb 18
|
|
—
|
AVOID
|
Mitchell Green
Founder of Lead Edge Capital
|
Green explicitly states we are in a "Giant AI CapEx Bubble" and that the amount of money being spent on infrastructure is "mind-boggling" relative to current revenue. Overbuilding is rampant. Similar to the telecom fiber bubble, capacity is being built that may not be utilized immediately, leading to massive depreciation cycles that will hurt the owners of this hardware/infrastructure. AVOID. The risk/reward for pure-play infrastructure build-out is skewed to the downside if utilization lags. If AI adoption accelerates exponentially (AGI), the demand for compute could outstrip even this massive build-out. |
Bloomberg Markets
Bloomberg Surveillance 2/18/2026
|
|
Feb 17
|
|
—
|
LONG
|
Thread Guy
Crypto influencer, independent
|
"Anything energy compute bottleneck I think goes up only forever." The US government's commitment to winning the AI race requires massive physical infrastructure build-outs. The bottleneck is no longer code, but the electricity and processing power required to run the models. Long the infrastructure layer supporting AI. Regulatory hurdles for new energy projects or hardware supply chain disruptions. |
Thread Guy
China is DOMINATING the US in EVERY statistic...
|
|
Feb 17
|
|
—
|
LONG
|
Thread Guy
Crypto influencer, independent
|
"Compute is just going to be the most in demand vertical forever for the end of time. You will never be able to get enough of it forever to run this [__] forever." If 100% of code is written by AI, the constraint on software creation is no longer human labor but the processing power required to run the models (Opus 4.6/Codex). Demand for chips and data centers will become infinite as software production becomes frictionless. LONG the infrastructure and hardware powering the AI transition. Supply chain constraints or energy limitations preventing the scaling of compute. |
Thread Guy
Why AI Is Taking Over ALL Coding Jobs..
|
|
Feb 17
|
|
—
|
LONG
|
Thread Guy
Crypto influencer, independent
|
"Demand for compute is just the most up only J curve of all time. It might never stop... Whether you're being useful or not... You're burning the same [__] compute." The utility of the output doesn't matter for the hardware provider. HFTs, startups, and enterprises are "cooking tokens" at an exponential rate. This indiscriminate consumption ensures sustained, compounding revenue for the underlying infrastructure and chip providers. LONG. Physical power constraints (electricity) or a sudden burst in the AI valuation bubble. |
Thread Guy
This OpenAI Acquisition Changes Everything..
|
|
Feb 17
|
|
—
|
LONG
|
Justin Wheeler
CEO, Berkadia
|
"There seems to be just tons of demand for it... we think that the next five years there's not an over supply situation." Despite fears of AI efficiency reducing space needs, the physical infrastructure required for compute (Hyperscalers) continues to outpace supply. Long Data Center REITs and infrastructure providers. Exit risk/valuation concerns if liquidity dries up; long-term AI efficiency reducing physical footprint. |
CNBC
Property Play: Scenes from a CRE finance conf...
|
|
Feb 16
|
|
—
|
LONG
|
Thread Guy
Crypto influencer, independent
|
"All compute cost explode... demand for compute is just the most upon onlyly JC curve of all time." If 100% of code is now written by AI (as claimed by TG's sources), the volume of software being created will explode exponentially. This software requires agents to run, and agents require massive compute. The demand for tokens and processing power is effectively infinite. LONG the entire compute stack (Hardware, Chips, Energy). Physical constraints on energy or chip manufacturing capacity. |
Thread Guy
LIVE: OPENAI BOUGHT OPENCLAW! ANTHROPIC IS CO...
|
|
Feb 14
|
|
—
|
LONG
|
Steve Rattner
Economic Analyst / CEO of Willett Advisors
|
"There's obviously a huge boom in construction going on in data centers... but I don't have any sign... that the number of car plants in this country is increasing." While general manufacturing (factories) is in secular decline, the build-out of AI/Cloud infrastructure is the singular bright spot in US capital investment. This concentrates demand on construction machinery, power management, and HVAC specific to data centers. LONG. This is the only sub-sector of "manufacturing/construction" with actual momentum. Regulatory pauses on power consumption or AI capex spending cuts. |
Bloomberg Markets
Is Trump’s Manufacturing Comeback Real?
|
|
Feb 14
|
|
—
|
LONG
|
Steve Rattner
Economic Analyst / CEO of Willett Advisors
|
While general manufacturing is struggling, Rattner states there is "obviously a huge boom in construction going on in data centers." He also cites the CHIPS Act as a success with factories building in Arizona. Capital expenditure is decisively shifting from traditional industrial plants to digital infrastructure. Companies providing the physical infrastructure (power/cooling like Vertiv) and the chips (Semiconductors) are the sole beneficiaries of this "manufacturing" spend. LONG the picks and shovels of the data center buildout. Overbuild/capacity glut or energy supply constraints. |
Bloomberg Markets
Wall Street Week | Rattner on Manufacturing, ...
|
|
Feb 13
|
|
—
|
LONG
|
Jim Cantrell
CEO, Phantom Space (Co-Founder SpaceX)
|
"What we hear from the HYPERSCALERS is they're looking for unique data... that unique data resides in space... We think that's one of the killer apps [AI]." The AI trade is evolving from "chip manufacturing" to "data acquisition." As terrestrial data becomes commoditized or exhausted, the premium shifts to companies that can harvest and process unique physical-world data from orbit. Long the AI value chain that extends into physical infrastructure and data acquisition. Overvaluation in the AI sector; the speaker notes the industry recently went through an "asset bubble" in 2021/2022 that has since burst and is recovering. |
Bloomberg Markets
Building Data Centers in Space
|
|
Feb 12
|
|
—
|
LONG
|
Alex Bores
NY State Assembly Member
|
Bores highlights New York State's $400 million investment to build its own "compute cluster" (Empire AI) and explicitly states, "We should be doing that same thing in investing in the capacity at the federal level." If Bores' plan for a federal version of the "Raise Act" includes the government purchasing its own compute capacity to expedite research and create a public option, the U.S. government becomes a massive direct buyer of AI hardware (GPUs). This adds a new, sovereign layer of demand for chipmakers (NVDA) beyond private sector capex. LONG hardware/infrastructure providers as government-sovereign AI spending ramps up. Legislative gridlock; federal budget constraints. |
CNBC
NY Assemblyman Alex Bores on AI regulation: N...
|
|
Feb 12
|
|
—
|
LONG
|
Raj Subramaniam
CEO and President of FedEx
|
"We are now heavily focused in on high value industry verticals... whether it is healthcare... whether it's the data center business... can you imagine you know when all this investments that is happening in data center there's a lot of supply chain elements involved in moving those goods." FedEx possesses real-time data on the physical economy. If they are aggressively pivoting resources to service Data Centers and Healthcare, it confirms that the AI capex cycle (moving GPUs, servers, cooling units) and medical logistics are not just digital narratives but are generating sustained physical volumes. This validates the "picks and shovels" thesis for these sectors. LONG. Follow the logistics capital; physical volume confirms the digital trend. AI capex spending slowdown; regulatory changes in healthcare logistics. |
CNBC
FedEx CEO Raj Subramaniam: We are undergoing ...
|
|
Feb 12
|
|
—
|
WATCH
|
Josh Shapiro
Governor of Pennsylvania
|
Shapiro outlines a principle where "If you're a big hyperscaler... you've got to be able to bring and pay for your own energy" via a "secondary auction." This shifts the cost burden of new generation explicitly onto Data Centers rather than spreading it across all ratepayers. While it ensures they get power (positive for growth), it likely increases their specific operating costs (negative for margins) compared to a subsidized model. Watch for the implementation of "secondary auctions" which could formalize higher energy costs for tech giants. If the secondary market fails to develop, data centers may face power shortages in the PJM region. |
CNBC
Gov. Josh Shapiro: PJM model of raising price...
|
|
Feb 11
|
|
—
|
LONG
|
Donald Trump
President of the United States
|
"Coal is also critical to... artificial intelligence... It's incredible what's happening with coal." The President views coal as the primary solution to the "energy cliff" facing AI data centers. This implies that regulatory hurdles for powering new data centers will be removed *if* they utilize fossil fuel baseloads. This removes the power-constraint bottleneck for AI scaling. LONG AI infrastructure and data centers, as energy constraints are being legislated away via coal deregulation. Tech companies (Amazon, Microsoft, Google) refusing coal power due to internal ESG mandates despite government pressure. |
CNBC
President Trump participates in an event on c...
|
|
Feb 11
|
|
—
|
LONG
|
Nick Carter
General Partner, Castle Island Ventures
|
Nick states, "I am not worried in the slightest about 600 billion of capex... I think this is bigger than the industrial revolution." He explicitly prefers owning the "company making this happen" (infrastructure) over the model companies. AI capability is improving on a "super exponential" curve. Regardless of which model wins (OpenAI vs. Anthropic vs. Google), they all require massive compute and energy. The infrastructure layer is the "pick and shovel" play that hedges against model obsolescence. Long the physical infrastructure powering AI. Overbuilding of capacity if AI monetization slows down. |
Unchained (Chopping Block)
Would BlackRock Try to Save Bitcoin From the ...
|
|
Feb 09
|
|
—
|
WATCH
|
Ali Ghodsi
CEO, Databricks
|
There is unprecedented capital expenditure ($50B–$100B) flowing into hardware, data centers, and energy. While the AI trend is real, the current build-out creates a risk of "overbuilding." The market is pricing in perfection, but physical constraints (energy) and ROI questions remain. The sheer volume of capital chasing hardware, combined with "circular" funding deals in the AI startup ecosystem, mirrors the pre-crash vibes of 2000. If AI adoption accelerates faster than hardware supply, these stocks will continue to run despite valuation concerns. |
CNBC
Under the hood of the AI economy: Databricks ...
|