Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
1 selected
Loading...
0 selected
Loading...
0 selected
0 selected
All post types✓
Portfolio updates✓
Stock lists✓
Research✓
News✓
All Content✓
Source feeds
Buzzberg Top 50
✓
All market capsNo capitalization filter✓
200 B and aboveMega✓
10 B to 200 BLarge✓
2 B to 10 BMid✓
0 to 2 BSmall✓
Custom
Enter market cap range in B USD
All directions✓
▲ Long✓
▼ Short✓
⛔ Avoid✓
✂ Close✓
◦ Others✓
Any score✓
LOW+✓
MED+✓
HIGH✓
14:00
Sep 02
Sep 02
SemiAnalysis
2d
SKYY 1ST
Neoclouds
AI compute providers
CIBR
▾
HIGH
Hyperscalers safer than most neoclouds.
Hyperscalers set the enterprise security bar, while neoclouds have huge security variability; only certain neoclouds are approaching hyperscaler standards, and startups spending large amounts of VC on GPUs face underappreciated counterparty risk if they choose the wrong providers.
SKYY LONG
Neoclouds WATCH
Compute access is frontier moat.
Access to frontier models and compute is becoming more important than access to top human experts for winning in math, cybersecurity, software engineering, trading, drug discovery, and autonomous vehicles, making AI compute/model access a key bottleneck and demand driver.
AI compute providers LONG
AI cyber impact slower than expected.
Public data do not yet show a rocket ship of AI-driven CVE discoveries or security patches; open source repos show more code churn and slightly more security-tagged changes, but no clear step-change, suggesting AI cyber impact is arriving slower than security company commentary implies.
CIBR WATCH
HIGH
15:00
Aug 29
Aug 29
SemiAnalysis
6d
005930.KS 1ST
MU 1ST
AMD 1ST
NVDA
000660.KS 1ST
▾
HIGH
Samsung leads HBM4; SK hynix, Micron lag.
Samsung has emerged as the HBM4 leader because its DRAM dies are built on a more advanced 1C process and its base die uses Samsung Foundry SF4 4nm, while SK hynix and Micron are on older processes and have faced HBM4 delays or redesigns. Samsung's HBM4 supplies OpenAI Jalapeño/Broadcom with 15.4 TB/s bandwidth, higher than expected NVIDIA Rubin HBM4, so Samsung is the HBM4 winner while SK hynix and Micron lag.
005930.KS LONG
MU AVOID
000660.KS AVOID
AI-assisted porting reduces AMD software disadvantage.
Anthropic is bringing in AMD as a hardware provider because agentic AI programming lets labs bypass AMD's software stack challenges. This suggests AI-assisted kernel development is eroding NVIDIA's software moat and expanding the addressable market for AMD accelerators.
AMD LONG
OpenAI Jalapeño threatens Nvidia's CUDA moat.
OpenAI's Jalapeño is the first non-NVIDIA, non-AMD accelerator to beat NVIDIA's GB300 and Vera Rubin on public inference benchmarks, winning on both throughput per megawatt and interactivity. This signals that the CUDA moat is eroding as AI-assisted programming and vertically integrated ASICs reduce NVIDIA's software lock-in, though NVIDIA's broader ecosystem and supply chain still make it valuable.
NVDA WATCH
Custom ASIC shift benefits Broadcom's design margins.
Custom in-house AI accelerators can lower hyperscaler chip costs because the buyer pays Broadcom's design margin only, rather than NVIDIA's or Broadcom-plus-Google-TPU margins. OpenAI's Jalapeño TCO advantage on tokens per dollar versus GB300 and Vera Rubin supports this custom ASIC economics, making Broadcom a key beneficiary of hyperscaler silicon programs.
AVGO LONG
HIGH
11:00
Aug 20
Aug 20
SemiAnalysis
15d
PJM existing power generators
Battery storage
▾
MED
Existing PJM generators profit from scarcity pricing.
PJM's capacity auctions are creating scarcity pricing because supply is constrained by short lead times and interconnection delays while demand is inflated by modeling errors. Existing generators are direct beneficiaries: they are currently receiving windfalls from constrained supply and higher capacity prices, and those high-priced capacity contracts last until roughly mid-2029.
PJM existing power generators LONG
Battery storage is a major grid opportunity.
Better deployment of batteries is described as a very large opportunity to bring down electricity prices and improve grid utilization and flexibility. Battery storage is one of the main ways the system can capture efficiency gains as load grows, rather than relying only on scarcity-priced capacity.
Battery storage LONG
MED
15:00
Aug 17
Aug 17
SemiAnalysis
18d
AVGO 1ST
SMH
NVDA 1ST
▾
HIGH
Nvidia and Broadcom capture bulk AI cash.
Alternative AI accelerators and startup chip companies may win some real orders, but their volumes are tiny compared with Nvidia and Google TPUs, and many startup orders are just nebulous LOIs. As long as demand outstrips production and incumbents keep improving, the bulk of AI accelerator revenue and cash flows will go to incumbents such as Nvidia and Broadcom.
AVGO LONG
NVDA LONG
AI compute demand outstrips supply, prices rise.
Dylan argues AI compute demand is still outstripping supply and the gap is widening, so the price of compute continues to rise even as unit costs may fall; he frames this as a structural bull case for the semiconductor and compute complex.
SMH LONG
HIGH
14:01
Aug 09
Aug 09
SemiAnalysis
26d
NVDA 1ST
MSFT 1ST
GOOGL 1ST
▾
HIGH
SpaceX bet on Nvidia drives huge demand.
SpaceX's decision to go all-in on Nvidia GPUs as their exclusive provider for the 10 GW AI data center buildout is a massive validation for Nvidia. SpaceX's enormous capex and the likelihood that Nvidia will provide vendor financing will lock in high-volume GPU orders, significantly boosting Nvidia's revenue and ecosystem lock-in even as competitors push alternative chips.
NVDA LONG
Microsoft fills capacity gap with SpaceX compute.
Microsoft, with exclusive access to OpenAI's frontier models, can monetize AI tokens at around $100 million per megawatt per year. Their current data center pipeline leaves a capacity gap in late 2026/early 2027. SpaceX's on-demand, instantly available compute with a 90-day cancellation clause perfectly fills that gap, enabling Microsoft to capture huge AI revenue growth without long-term balance-sheet risk.
MSFT LONG
Google falls behind in AI frontier race.
Google is falling behind in the AI frontier race. Key researchers like Jeff Dean are leaving to raise external funds, signaling a loss of talent. Google's focus is not competitive in coding and AGI, and their massive capex is not translating into a leading position, making the stock unattractive as an AI play.
GOOGL AVOID
HIGH
22:00
Aug 07
Aug 07
SemiAnalysis
28d
GOOGL 1ST
DRAM 1ST
▾
HIGH
Google stock good short/medium term.
Google has an 'L culture' that fails to execute internally, acquiring most innovations, but its search monopoly will be protected and possibly anointed by the government, making the stock attractive in the short and medium term despite the loss of top AI talent and eventual long-term disruption risk (similar to Bell Labs).
GOOGL LONG
Memory supply is tight, bullish memory.
Memory (DRAM/NAND) is in severe shortage, crimping product output in Taiwan; Apple has N2 chips sitting idle because it cannot get memory for phones, and the constraint is so acute that expanding memory production is the fastest route to revenue, strongly benefiting memory manufacturers.
DRAM LONG
HIGH
20:59
Jul 29
Jul 29
SemiAnalysis
1mo
000660.KS 1ST
MU 1ST
▾
HIGH
Memory cycle stronger, buy on weakness.
SK Hynix missed consensus due to shifting more to LTAs, but the memory cycle is bigger, longer, and stronger than past cycles. Demand still exceeds supply, and the recent selloff is driven by technicals and leverage unwinds, creating a buying opportunity in memory stocks.
000660.KS LONG
MU LONG
HIGH
22:00
Jul 23
Jul 23
SemiAnalysis
1mo
NVDA 1ST
ANET 1ST
GPUS 1ST
▾
HIGH
Backstops turn NVIDIA hardware into recurring revenue.
NVIDIA is leveraging its GPU backstop program to enforce strict ecosystem lock-in among NeoClouds. To receive a backstop and NVIDIA Certified Partner (NCP) status, providers must commit to purchasing NVIDIA's Spectrum-X switches and branded LinkX transceivers, completely locking out competing networking hardware from companies like Arista Networks.
NVDA LONG
NVIDIA backstops force exclusive use of its networking.
NVIDIA is leveraging its GPU backstop program to enforce strict ecosystem lock-in among NeoClouds. To receive a backstop and NVIDIA Certified Partner (NCP) status, providers must commit to purchasing NVIDIA's Spectrum-X switches and branded LinkX transceivers, completely locking out competing networking hardware from companies like Arista Networks.
ANET AVOID
AI training compute demand will remain robust.
The market incorrectly assumes that AI compute demand will eventually shift entirely to inference, halting training investments. In reality, frontier AI labs will continue to invest as much capital as possible into training increasingly massive models, ensuring that demand for training GPUs remains structurally robust.
GPUS LONG
HIGH
01:00
Jul 22
Jul 22
SemiAnalysis
1mo
MSFT 1ST
GOOGL 1ST
META 1ST
AMZN
▾
HIGH
Enterprise token spend flows to AWS, Azure.
Token-as-a-service through hyperscalers like AWS Bedrock and Azure Foundry is a very popular and attractive channel for enterprises, especially regulated industries like financial services, because they can buy AI tokens through existing cloud vendor relationships and burn down committed credits. This should drive significant growth for AWS and Azure as token spending shifts towards enterprise.
MSFT LONG
AMZN LONG
Google AI lags, locked-in TPU deals.
Google's Gemini models are clearly in fifth place and likely to stay there because they are not true frontier. Google has signed long-term TPU deals without clawback clauses, indicating a lack of conviction in its own ability to build AGI/RSI, unlike competitors who structure deals to retain flexibility to claw back compute for frontier training.
GOOGL AVOID
Meta compute strategy de-risks capex spend.
Meta's compute neocloud strategy gives it a backstop for its massive AI capex. If its own model (MSL) doesn't succeed, Meta can lease GPU capacity to AI labs at premium rates, providing investors confidence and enabling continued aggressive capex spending in 2027 and beyond.
META LONG
HIGH
14:00
Jul 20
Jul 20
SemiAnalysis
1mo
0981.HK 1ST
▾
MED
SMIC progress makes China a fab power.
SMIC's latest DUV node, N+3, has achieved EUV-class density without EUV, demonstrating real progress. Although power and performance still lag the leading edge, this shows China is becoming a serious fab player. SMIC is being directed to license its N+2 and N+3 processes to other domestic fabs. If this learning spreads into AI accelerators and other chips, the choke point moves from a single sanctionable fab to an entire ecosystem, making China less dependent on TSMC and harder to contain. China doesn't need to beat TSMC; it only needs to be good enough to not need TSMC at all.
0981.HK LONG
MED
01:08
Jul 18
Jul 18
SemiAnalysis
1mo
NVDA 1ST
AMD 1ST
GOOGL 1ST
▾
MED
Frontier models demand latest Nvidia and AMD accelerators.
Serving large frontier models like Kimi K3 with 2.8 trillion parameters requires the latest and most powerful accelerators such as Nvidia B300/GB300 or AMD MI355X because the model does not fit on older B200 systems. This creates immediate demand for next-generation AI chips.
NVDA LONG
AMD LONG
Google embarrassed as AI leader.
Google is no longer among the top three AI model developers and has fallen behind Anthropic, OpenAI, and now Moonshot. This competitive slippage is so stark that Google should be 'incredibly embarrassed,' signaling loss of leadership in a critical technology area.
GOOGL AVOID
MED
21:00
Jul 16
Jul 16
SemiAnalysis
1mo
0981.HK 1ST
▾
MED
SMIC scales aggressively despite EUV restrictions
SMIC's N+3 process achieves extremely aggressive scaling without EUV, reaching parity with the world's leading fabs in some aspects while managing yield and performance through innovative process decisions. Export restrictions have forced SMIC to innovate, creating progress despite being unable to use EUV.
0981.HK LONG
MED
23:00
Jul 15
Jul 15
20:00
Jun 25
Jun 25
19:00
Jun 11
Jun 11
22:15
May 27
May 27
19:00
May 18
May 18
SemiAnalysis
3mo
SMH 1ST
▾
MED
Semiconductor stocks only go up.
Compute for AI is still a limited resource, driving sustained demand for semiconductor stocks. AI models are increasingly restrictive with tokens and downgrading to save compute, and the latest frontier models require massive compute. This backdrop supports continued upside for semiconductor stocks.
SMH LONG
MED
00:01
May 14
May 14
19:30
Apr 30
Apr 30
19:58
Apr 16
Apr 16
SemiAnalysis
4mo
HBM 1ST
NVDA 1ST
SOXX 1ST
▾
HIGH
Commodity LPDDR and substrates over HBM and advanced packaging.
Positron AI uses LPDDR memory and organic substrates instead of HBM and advanced packaging, avoiding supply chain constraints and achieving higher memory capacity for scalable AI inference with commodity technologies.
HBM AVOID
SOXX LONG
NVIDIA has poor matrix-vector performance for inference.
NVIDIA's GPUs have a worsening ratio of matrix-matrix to matrix-vector performance from Hopper to Blackwell, making them inefficient for AI inference workloads that rely heavily on matrix-vector operations, which Positron's architecture solves.
NVDA AVOID
HIGH
Choose Sources
Loading sources...