🌐 Module 8 · Context and the Future · Chapter 8.1 · 9 min read

The AI Chip Competition

NVIDIA, Intel, Mythic, Cerebras — and SIDRA's place.

Prerequisites

What you'll learn here

  • Name the major players in the global AI chip market
  • State each company's technical approach and market share
  • Identify market segmentation (datacenter, edge, mobile)
  • Summarize the US-China semiconductor race
  • State SIDRA's place in the global market and its differentiation

Hook: The Global AI Chip Race

As of 2024, the AI chip market is 50B/year∗∗,expectedtoreach∗∗50B/year**, expected to reach **200B+ by 2030.

Leader: NVIDIA (~80% share). Rivals: Intel, AMD, Google TPU, AWS Trainium, Mythic AI, Cerebras, SambaNova, Graphcore.

Where is SIDRA YILDIRIM? Niche segment: neuromorphic analog memristor. Same category as Mythic + Rain.

Intuition: 4 Categories

CategoryLeaderMarket shareApproach
GPU (large datacenter)NVIDIA H10080%Classic, FP16/8 digital
TPU (Google)TPU v55%Systolic-array digital
Wafer-scaleCerebras CS-22%Single-wafer chip
Neuromorphic analogMythic, Rain, SIDRA1%Memristor/flash analog

SIDRA is in the 1% niche but on a path to lead the neuromorphic analog category.

Formalism: Company-by-Company

L1 · Başlangıç

NVIDIA:

  • Founded 1993, GPU leader.
  • H100 (2022): 700 W, ~4 PFLOPS FP8.
  • B200 (2024): 1000 W, ~10 PFLOPS.
  • 2024 datacenter revenue: $48B.
  • AI’s “gold standard”.

Intel:

  • Habana Gaudi (Intel acquired 2019).
  • Gaudi 3 (2024): 600 W, 1.8 PFLOPS.
  • Small share (~2%).

Google TPU:

  • 2016 first version, 5th gen 2024.
  • Systolic-array digital.
  • Google datacenter only (no commercial sales).

Cerebras:

  • 2016 startup. Wafer-scale (one wafer = one chip).
  • CS-3 (2024): 850K cores, 125 PFLOPS.
  • $4B valuation.

Mythic AI:

  • Analog NAND-flash compute. 25 TOPS/W.
  • 2022 product, limited sales.
  • $200M private.

Rain AI:

  • Analog photonic. In development.
  • $80M private.

SIDRA YILDIRIM (Türkiye):

  • 2024 startup. Memristor analog CIM.
  • Y1 (2026): 30 TOPS @ 3W, 10 TOPS/W.
  • Y100 (2031): 3 PFLOPS @ 100W, 300 TOPS/W.
  • Center of Türkiye’s semiconductor strategy.
L2 · Tam

Market segmentation:

Datacenter training:

  • NVIDIA H100/B200, Google TPU, AMD MI300.
  • High budget, high power.
  • SIDRA absent (insufficient for training).

Datacenter inference:

  • NVIDIA L40, Mythic, SIDRA Y10+.
  • Low power, high throughput.
  • Growing segment ($20B in 2030).

Edge (mobile, IoT):

  • Apple Neural Engine, Qualcomm AI, NVIDIA Jetson.
  • 1-10 W budget.
  • SIDRA Y1 ideal here.

Embedded (automotive, defense):

  • ASELSAN, NVIDIA Drive, Qualcomm.
  • Strict certification, long lifetime.
  • SIDRA Y3+ candidate.

SIDRA market position:

Y1 → edge inference, embedded. Y10 → high-perf edge + datacenter inference. Y100 → datacenter standard inference.

Training segment a long-horizon Y1000.

Market size estimates:

Segment20242030SIDRA target
Datacenter training$30B$80B0%
Datacenter inference$5B$40B5% (Y100)
Edge$10B$50B2% (Y10)
Embedded$5B$30B3% (Y10+)
Total$50B$200B~$5B SIDRA

2.5% global = an extraordinary target for SIDRA. Realistic 0.5-1% = $1-2B/year by 2030. Still huge for Türkiye.

L3 · Derin

Geopolitics:

US-China semiconductor war:

  • US: advanced-chip exports to China blocked (post-2022).
  • China: building its own fabs (SMIC, Hua Hong).
  • Taiwan: TSMC has strategic importance.

ASML EUV:

  • Netherlands-based, sole supplier for 5 nm and below.
  • US pressure blocks sales to China.
  • A single EUV machine: $200M.

Türkiye’s position:

  • US ally but extra export restrictions (outside defense).
  • Not in the EU → ASML possibly accessible.
  • Trade with China unrestricted.
  • Middle position: access to both US and China tech.

SIDRA can turn this middle position into an advantage: domestic production + flexible international supply.

Strategic trend:

Semiconductors viewed as a strategic resource:

  • US CHIPS Act ($53B incentive 2022).
  • EU Chips Act (€43B 2023).
  • China Big Fund (30B2014,30B 2014, 40B 2024).
  • Türkiye: no big incentive yet. SIDRA can fill that gap.

Semiconductor sovereignty:

Semiconductors are strategic — like oil or water. Nation-state infrastructure.

SIDRA gives Türkiye the first concrete step toward that infrastructure.

Numerical comparison vs rivals:

CompanyTOPS/WMarketTürkiye
NVIDIA H1006$48BImported
Mythic25$50MImported
SIDRA Y10100$500M (target)Domestic

Y10 is both technically superior and domestic.

Experiment: SIDRA SWOT

Strengths:

  • Memristor know-how (UNAM 10+ years of research).
  • Türkiye semiconductor ecosystem support.
  • Low startup cost.

Weaknesses:

  • No market experience.
  • International sales channels missing.
  • TSMC dependency.

Opportunities:

  • Edge AI explosion ($50B by 2030).
  • Türkiye’s strategic position.
  • Open neuromorphic category (no clear leader yet).

Threats:

  • NVIDIA dominance.
  • China cheap alternatives.
  • Geopolitical risk.

Strategy: niche (neuromorphic edge inference) + strong Türkiye brand. Become a global player in 5-10 years.

Quick Quiz

1/6AI chip market leader?

Lab Exercise

SIDRA’s 5-year market strategy.

2026: Y1 product → Türkiye domestic (ASELSAN, universities). 2027: Expansion → Turkish startups (mobile apps, IoT). 2028: Europe export (DACH region). 2029: Y10 prototype → datacenter inference. 2030: Y10 serial production → global inference market.

End of 5 years: $1B/year revenue target. Extraordinary but realizable.

Cheat Sheet

  • AI chip market 2024: 50B;203050B; 2030 200B.
  • Leader: NVIDIA 80%.
  • Niche: neuromorphic analog (Mythic, Rain, SIDRA).
  • SIDRA target: $1-5B/year by 2030.
  • US-China race: Türkiye middle position.
  • State incentives: CHIPS Act, EU Chips Act, China Big Fund — none yet in Türkiye.

Vision: Global SIDRA

  • Y1: Türkiye domestic.
  • Y3: Europe.
  • Y10: Global inference market.
  • Y100: Datacenter standard.
  • Y1000: Among the semiconductor giants.

Türkiye becomes a global tech producer by 2035.

Further Reading