TickerTain
TickerTain
NewsroomShortsPortfolioConvergence
NewsroomShortsPortfolioConvergence
←
▶ 4:50 · $CBRS · The $100B Chip IPO Challenging Nvidia (Cerebras)
episode briefing
Michael Sikand

The $100B Chip IPO Challenging Nvidia (Cerebras)

2026-05-10 · 5 company · 3 thematic
sentiment
1 bull1 bear3 neu
speakers
michael sikand

Founder of Our Future, a business-media brand focused on the next generation of founders and operators. He also hosts interviews with entrepreneurs and executives.

quote
“All right, time to dive into the actual numbers in the filing. So, Cerebras revenue grew 76% from 290 million in 2024 to 510 million in 2025. And there's two segments. Hardware is 70% of sales. In this business, companies are buying the ch…”
—michael sikand
now playing · $CBRS
$CBRS···neutral· mediummichael sikand
Cerebras files for IPO with wafer-scale chip targeting AI inference dominance
Cerebras's wafer-scale engine replaces 32-64 Nvidia GPUs per chip, delivering 15x faster inference with lower power, but business depends on OpenAI $20B contract and UAE G42 for 8…
$OPENAIneutral· mediummichael sikand
OpenAI's $20B Cerebras contract and IPO success critical to Cerebras investment thesis
Cerebras's revenue could 8x to $4B/year from OpenAI partnership, but OpenAI lost $13.5B in H1 2025 and projects $665B losses through 2030; Anthropic is winning enterprise deals, m…
$AMZN···bullish· lowmichael sikand
AWS partnership could diversify Cerebras revenue away from UAE and OpenAI concentration
AWS is buying Cerebras systems upfront and renting compute to customers, which could broaden Cerebras's customer base beyond G42 and OpenAI.
$GROQbearish· mediummichael sikand
Nvidia acquires Groq, eliminating Cerebras's closest pure-play inference competitor
Nvidia's $20B acquisition of Groq brings infinite resources to the inference market, directly challenging Cerebras's speed advantage.
$NVDA···neutral· mediummichael sikand
Nvidia faces inference challenge as Cerebras offers 15x speed advantage with wafer-scale architecture
Nvidia's GPU clusters require complex networking and CUDA software, creating latency and power overhead for inference; Cerebras's single massive chip eliminates inter-chip communi…