Research from @creativestrat analyst-grade intelligence on the companies reshaping the technology landscape. We analyze the full stack—from silicon to software.
- Tweets 53
- Following 8
- Followers 1,061
- Likes 7
Expert Interview: From Gas to Gigawatts and What It Takes to Power AI
Latest report with companion research deep dives in CS Atlas.
thediligencestack.com/p/expe…
3
6
1,608
The Diligence Stack retweeted
All the research we do, all field notes, event coverage, exec interviews, data models, surveys, and so much more goes into CS Atlas. We do sooooo much research and analysis it just needed a home.
@DiligenceStack subs even get entry tier access :)
atlas.creativestrategies.com…
1
3
22
5,197
The Diligence Stack retweeted
New report @DiligenceStack with links to our expanded coverage of AI infra summit in CS Atlas for subscribers and clients.
The Future of Compute Is Fungible:
How greater choice in data center infrastructure changes the economics for operators and suppliers
thediligencestack.com/p/the-…
2
3
2,976
Optical demand does not become deployable supply when new capacity is installed. Laser die, precision packaging, qualification and field serviceability determine how quickly NPO and CPO can move into AI racks.
thediligencestack.com/p/opti…
1
2
273
Enterprise AI is moving into governed production one workflow at a time. The useful denominator is cost per completed task at required quality, including model use, implementation, review and rework.
thediligencestack.com/p/stat…
3
228
Inference architecture is a bet on where the bottleneck sits. Moving model state closer to compute can cut latency and energy, but the gain has to exceed the added software and operating cost.
thediligencestack.com/p/ais-…
1
166
Credo’s Next Phase of Growth
Why broader design ownership can help Credo win more customer business, and what that growth needs to earn.
With full deep research companion on AI infra interconnects in CS Atlas.
$CRDO
thediligencestack.com/p/cred…
1
2
7,175
Agentic AI’s Next Frontier: Cyber Defense
More capable models can help enterprises investigate threats and respond. The next challenge is deciding how much authority to give them.
thediligencestack.com/p/agen…
1
1
5
1,678
Custom Silicon 3.0: Growth and Competition Across Compute and Networking
Why the market is shifting from ASIC design wins to program responsibility across compute and networking
thediligencestack.com/p/cust…
1
4,202
CS Atlas makes the full Creative Strategies research archive searchable and source-linked. Subscribers can ask a question, see the research behind the answer, and move into the underlying reports and models.
thediligencestack.com/p/intr…
1
3
1,906
The GPU Tsunami: Testing, Testing.
If you know what you are looking for the signals in semiconductor testing present a signal for future design trends.
Our report.
thediligencestack.com/p/the-…
1
6,108
Chip shipments no longer capture semiconductor test demand. Larger accelerators occupy testers longer, while advanced packages add more places to check for failure—raising test work per sellable AI system.
thediligencestack.com/p/the-…
1
167
AI infrastructure grows at the pace of its slowest layer. Capacity added through 2027 still needs time to qualify and reach useful yield, pushing broader supply relief into 2028.
thediligencestack.com/p/2027…
1
4
1,458
AI projects need security approval before they reach production. That gives existing data and identity vendors a path to more spending, with paid adoption and renewals testing the revenue case.
thediligencestack.com/p/cybe…
1
195
CXL lets hyperscalers reuse qualified DDR4 from retired servers. Keeping memory in service through a CPU refresh can improve the payback on the hardware needed to attach it.
thediligencestack.com/p/brea…
1
191
More laser capacity can leave optics bottlenecked in packaging. Each assembly still needs precise alignment and must pass reliability tests before it becomes usable supply for AI racks.
thediligencestack.com/p/opti…
1
162
When power is fixed, the winning AI platform is the one that converts each active megawatt into the most gross profit. Useful output and utilization determine whether higher hardware cost earns its way back.
thediligencestack.com/p/giga…
1
197
Early AI forecasts counted accelerators. The market expanded faster because usable compute requires the full rack, while constraints raised semiconductor value per device far faster than unit shipments.
thediligencestack.com/p/in-2…
1
155
AI demand held. What changed was the constraint map: power availability and advanced packaging now determine how quickly accelerator supply becomes billable capacity.
thediligencestack.com/p/the-…
1
207
AI’s Inference Era of Ferment
Hot Chips showed an industry aligned on the inference problem and divided over the system that will solve it.
thediligencestack.com/p/ais-…
1
3
7
1,907