Small Cap Snipa
@SmallCapSnipa
Stock market research. AI / data centers / power infra / cloud compute. NFA
318 Following 35.3K Followers
J.P. Morgan sees hyperscaler capex reaching $1.7T in 2027 and eclipsing $2T in 2028
Up from $265B in 2024 a near 8x in four years.
$AMZN $GOOGL $META $MSFT $SPCX $ORCL
$IREN received a 90% prepayment on GPU capex for a five-year contract yielding $22M/MW
$CRWV $NBIS $SPCX
Bernstein’s $IREN model assumes a ~2 GW buildout by end of 2028 and AI cloud generating ~$16.8B in revenue by CY30E
Bernstein estimates OpenAI’s infrastructure spend will top out at $184B in 2031
The model estimates roughly ~$1.25-1.30T in AI infrastructure spend from 2025-2035
$NVDA $MSFT $GOOGL $AMZN $ORCL $CRWV
Show more
Muzuho sees Robotics TAM growing from $66B in 2025 to $427B by 2035 👀
Piper Sandler picks $HUT as highest upside HPC/Bitcoin miner:
HUT 8 likely offers highest upside from here. HUT's deals encompassing 949 MW Critical IT across its Beacon Point (704 MW with Nvidia) and River Bend (245 MW with Fluidstack/Anthropic/Google) campuses are proof points of the company's ability to execute. The contracted MWs come with expected annualized NOI of over $1.75B as HUT continues to secure the most favorable economics on its deals amongst its peers. With the company currently trading at an implied >9% cap rate on contracted power, we believe the company is undervalued. The quality & pace of HUT’s tenant wins strengthen our conviction that HUT is well positioned to secure future leasing commitments.
Show more
Same expert on why most startups and AI labs “hate” the hyperscalers and prefer neoclouds:
“The neoclouds instead, they operate in a completely different approach. The hyperscalers, it was a top-down approach. It means that from the platform and the ecosystem, they built then, in this case, the cloud computing clusters, and then they integrate it in their own platforms, while instead for the neoclouds, instead it's a bottom-up system. The neoclouds, we started with infrastructure, and then from the infrastructure, because we don't have anything at all as like the hyperscalers, they started to build what the AI natives wanted.”
$MSFT $META $SPCX $NBIS $CRWV $IREN
Show more
Expert call: Former global strategic partnership lead at NVIDIA
The biggest threat for Neoclouds and Bitcoin miners is $SPCX & $META dumping large amounts of available bare-metal GPU capacity into the market right now.
• Meta has ~1.3M GPUs with over 40% capacity underutilized
• Meta and SpaceX are signing bare-metal deals that can be accessed overnight (at a premium)
• That capacity and those deals would otherwise have gone to neoclouds
$NBIS $CRWV $IREN $KEEL
Show more
This looks awfully similar to what I posted yesterday, this time coming from the former global strategic partnership lead at Nvidia.
Multiple claims made here:
1. Building a commercial cloud takes at least 12 months, meaning massive quantities won't be online until 2028. A basic functional commercial cloud will still take at least 9 months to deploy.
2. Meta has roughly 1.3 million GPUs and can offload at least 600,000 older units directly to the market (such as H100, H200, B200, and B300) as it transitions to Vera GB300. Despite rising demand for models like Llama and Muse, over 40% of this capacity remains underutilized, posing a threat to the market.
3. Meta and SpaceX are providing immediate bare metal capacity deals, renting out clusters as large as 300,000 GPUs. While expensive, this instantly available supply takes high-value contracts (worth ~$20B–$30B from SpaceX and Meta) away from neoclouds.
4. Less specialized players and Bitcoin miners (e.g., Keel Infrastructure) face major hurdles because hyperscalers are favoring established neoclouds, Meta, or SpaceX. Without a strong anchor tenant to supply capital, it will be difficult for them to enter or grow in the market.
$SPCX $META $NBIS $IREN $CRWV
Show more
Expert call: Former global strategic partnership lead at NVIDIA
The biggest threat for Neoclouds and Bitcoin miners is $SPCX & $META dumping large amounts of available bare-metal GPU capacity into the market right now.
• Meta has ~1.3M GPUs with over 40% capacity underutilized
• Meta and SpaceX are signing bare-metal deals that can be accessed overnight (at a premium)
• That capacity and those deals would otherwise have gone to neoclouds
$NBIS $CRWV $IREN $KEEL
Show more
Jeffries: summary of changes HPC price targets
“We lower PTs across the group to reflect a rising cost of debt and now apply lower terminal value assumptions across our models.”
• $CIFR: $12 from $25
• $CORZ: $16 from $40
• $HUT: $90 from $140
• $RIOT: $13 from $25
• $WULF: $15 from $25
• $GLXY: $29 from $43
Show more
NEOCLOUD & AI DATA CENTER DELAYS
There have been several negative headlines in the broader AI infrastructure space (among both clouds and data center operators) including cancelations and power delays.
Show more
Jeffries top picks in HPC: $CIFR & $CORZ
Cipher PT: cut to $12 from $25 (-52%)
Corz PT: cut to $16 from $40 (- 60%)
Expert call: Former global strategic partnership lead at NVIDIA
The biggest threat for Neoclouds and Bitcoin miners is $SPCX & $META dumping large amounts of available bare-metal GPU capacity into the market right now.
• Meta has ~1.3M GPUs with over 40% capacity underutilized
• Meta and SpaceX are signing bare-metal deals that can be accessed overnight (at a premium)
• That capacity and those deals would otherwise have gone to neoclouds
$NBIS $CRWV $IREN $KEEL
Show more
Morgan Stanley: Global data center model implies over 300GW of capacity by 2030: Leveraging a bottom-up build by operator, along with a top-down total capacity forecast, we now expect a 23% CAGR in the growth of data center IT capacity between 2026 and 2030.
We expect ~180 GW of IT capacity to come online by 2030 in our base case, for a total of 316 GW (IT).
$AMZN $META $SPCX $MSFT $GOOGL $NVDA $CRWV
Show more
Bernstein on $INTC: “Q426 outlook however was (in a word) terrible.” PT cut to $75 from $110
HYPERSCALER CAPEX Q3 2026:
Microsoft lowered its CY26 outlook to ~$170B from ~$175B, attributing the Q3 shortfall to the timing of delivery of third-party capacity.
Google and Meta also came in light this quarter while Amazon beat expectations.
$GOOGL $META $AMZN $MSFT
Show more
BERNSTEIN: NEOCLOUD FINANCING HAS BECOME MORE RESTRICTIVE
The market has shifted from 100% guarantees to ~80% of construction debt rather than full obligation. Financing is still available, just tighter and more expensive than a year ago.
$CRWV $NBIS $IREN $DOCN $WYFI $SPCX $NVDA
Show more
$NBIS Expert Takeaways:
• Nebius enforces 60-65% upfront contract payments
• Short-term spot capacity generates $40-50M/MW
• Sort-term deals have potential to reach $60M/MW after full 20% auction-based price uplift
Show more
Expert Takeaways: Nebius
> Nebius operates an auction-driven cluster-access pricing model that drives up pricing levels by ~20% through competitive bidding.
> Nebius requires 60–65% upfront contract payments, providing strong protection compared to debt-heavy peers like CoreWeave (facing up to 15% interest on new debt).
> Short-term spot capacity fetches $40–50M ACV per megawatt (potentially up to $60M with auction uplift) vs. $20–25M for medium-term deals.
> Tight supply for latest-generation GPUs through at least mid-2028 supports reserving 2027 capacity for high-margin spot workloads.
> Meta and SpaceX control >1M GPUs and plan to deploy 25k–50k units monthly, pulling bare-metal contracts away from neoclouds.
> Open-source models deliver comparable results at 70–80% lower cost, eroding OpenAI and Anthropic enterprise revenue.
> Gartner projects >80% of independent neoclouds will fail by 2030 due to a K-shaped market favoring GB300/Vera-class hardware over older GPUs.
> Nebius is insulated from cascading neocloud defaults because its core clients (Microsoft, Meta) are AA/AAA-rated, unlike BB-rated frontier labs.
> High-end hardware (GB200/GB300) is shifting to inference workloads, which offer 4–8× higher margins than training.
> Micron predicts HBM3/HBM4 memory shortages beyond 2028, while Microsoft’s 300,000 internal Maia ASICs displace demand for 300,000+ Nvidia GPUs.
$NBIS $GOOGL $META $MSFT $AMZN
Show more
Morgan Stanley expects a 23% CAGR in global data center capacity supply, reaching 316 GW by 2030
$NVDA $SPCX $AMZN $GOOGL $MSFT $NBIS
Cramer says losing money buys $IREN & $NBIS:
“We don’t want IREN. There’s this thing called losing money. Losing money buys IREN and it buys Nebius.”
Jim Cramer on $RDDT exactly two months ago:
“I think Reddit is such a buy at $166 it’s ridiculous.”
Check the chart today.
