Tessara

TESSARA

FRI 13 MAR 2026 · 22:55 GMT+0

LIVE

My Tessara

Daily Chain Radar Watchlist

Intelligence

Signals Narratives

Sectors

Compute Memory Energy Models

Data

Companies Innovation Regulatory Methodology

IndicesCTI—MTI—PTI—

No dataSign in

Terms Privacy Refund policy

Narratives

Important narratives tracked by our Tessara intel engine. Each is scored by momentum (how fast it's gaining traction) and conviction (strength of evidence). The composite score is their average.

MOVED

Tessara

Narratives4 active

Important narratives tracked by our Tessara intel engine. Each is scored by momentum (how fast it's gaining traction) and conviction (strength of evidence). The composite score is their average.

MOVED

HIGH CONVICTION4

EmergingContestedHighAgents

AI agents need payment rails — Mastercard and Google own the chokepoint

Verifiable authorization infrastructure for AI agent transactions will be required by >3 major card networks and embedded in >5 enterprise software platforms by Q4 2026, making Mastercard and Google the default trust layer for autonomous commerce and locking out fintech challengers without similar verification schemes.

+44 −25 from 64

AlphabetAlibabaOpenAI

—8W TREND7h ago

DevelopingHighCompute

Custom 2nm SoCs are real — Broadcom and TSMC displace merchant silicon for hyperscale inference

Broadcom's 2nm custom SoC shipments to Fujitsu mark the beginning of a shift where hyperscalers and Tier-1 enterprises with >$1B annual compute spend will prefer custom silicon over merchant GPUs for inference by Q4 2026, measurable when custom ASIC inference share exceeds 20% of new hyperscaler inference capacity additions.

+56 −14 from 57

BroadcomNVIDIAUnitree Robotics

DevelopingContestedHighCompute

Hyperscaler GPU hoarding will force 40% of mid-tier AI startups to exit by Q2 2026

Hyperscaler GPU pre-commitments have exceeded $60B as of early 2026, extending non-hyperscaler H100/H200 lead times to 18+ months through Q2 2026 — falsified if lead times drop below 12 weeks before that date.

+77 −46 from 106

NVIDIAMeta PlatformsAMD

DevelopingHighComputeMemory

HBM3e pricing stays 40% above DDR5 through 2026 — inference API margins get crushed

HBM supply constraints will keep HBM3e spot pricing 40%+ above DDR5 through Q4 2026, compressing gross margins at inference API providers with <$1B revenue below 40% — falsified if HBM3e/DDR5 price gap closes below 20% before Q4 2026.

+69 −20 from 76

SK HynixNVIDIAMicron Technology