[SLIDE//01]
DOC: MONOLITH-SA-2026-V3.2 TIMESTAMP: 2026-06-18T00:00:00Z CLASSIFICATION: CONFIDENTIAL STATUS: SERIES-A-RAISE
MONO
LITH
Bare-metal compute infrastructure for the next decade of
AI workloads. No abstraction layers. No hyperscaler tax. No excuses.
$4.2M ARR
94% Gross Margin
11ms p99 Latency
3× YoY Growth
$18M Raise Target
MONOLITH TECHNOLOGIES INC.
SAN FRANCISCO / AUSTIN / AMSTERDAM
V3.2 // SERIES A // JUNE 2026
[SLIDE//02]
[PROBLEM_STATEMENT]

HYPERSCALERS OWN YOUR
INFRASTRUCTURE.

THREE STRUCTURAL PROBLEMS BLOCKING THE NEXT WAVE OF COMPUTE.
01
COST CAPTURE
72%
Of cloud compute spend is captured by AWS, Azure, and GCP at margins that leave operators paying 4–6× bare-metal cost equivalents. For AI inference workloads, the effective markup exceeds 9×.
02
PERFORMANCE CEILING
180ms
Median p99 on managed GPU inference across hyperscalers. Virtualization overhead, shared tenancy, and cold-start penalties destroy the latency profile that real-time AI products demand. MONOLITH delivers 11ms.
03
LOCK-IN PENALTY
$240K
Average egress and migration cost when an enterprise tries to move off a hyperscaler. Data gravity, proprietary APIs, and punitive egress pricing trap customers in perpetual contracts at captive rates.
THE WINDOW
AI inference demand is growing 400% YoY. Hyperscalers are supply-constrained. Lead times for H100 clusters exceed 9 months on AWS. The gap between demand and affordable supply is the largest infrastructure opportunity in a decade.

MONOLITH is the wedge.
[SLIDE//03]
[SOLUTION]

BARE METAL. ZERO
HYPERSCALER TAX.

[CORE_PLATFORM]
MONOLITH
COMPUTE FABRIC
Owned, collocated bare-metal GPU clusters across 8 Tier-3 data centers. Zero-overhead scheduling via our open-source Slab runtime — no hypervisor, no shared tenancy, no abstraction tax. Customers get the full silicon, billed per GPU-second at 70% below hyperscaler list price.
[KEY_DIFFERENTIATORS]
→11ms p99 guaranteed by SLA or credits issued automatically
→99.995% availability across cross-DC redundant pairs
→NVIDIA H100/H200 + custom InfiniBand topology — 3.2 Tbps inter-node
→Full data sovereignty — no telemetry, no co-mingling, HIPAA + SOC 2 Type II
[SLAB_RUNTIME]
OPEN-SOURCE
KERNEL SCHEDULER
Slab bypasses the Linux scheduler for GPU workloads, achieving near-zero context-switch overhead. 14K GitHub stars. 60+ contributors. Adopted by 3 of the top 10 AI labs.
[NETWORKING]
PRIVATE BACKBONE
INTERCONNECT
Dedicated dark-fiber routes between all 8 DCs. Customers transfer model weights and training checkpoints at 100Gbps with zero egress fees. No AWS PrivateLink required.
[SLIDE//04]
[PROOF_OF_PERFORMANCE]
11ms
P99 INFERENCE LATENCY — PRODUCTION MEDIAN // ALL CUSTOMERS // Q1 2026
vs. 180ms AWS SageMaker median
vs. 220ms GCP Vertex AI median
vs. 195ms Azure ML median

MONOLITH: 16× faster at p99.
$4.2M
Annual Recurring
Revenue
↑ 312% YoY
94%
Gross Margin
Q1 2026
↑ from 88% in 2025
148%
Net Revenue
Retention
↑ expansion MRR
6mo
CAC Payback
Period
↓ from 14mo in 2024
99.996%
Uptime TTM
Across all DCs
21 min downtime total
[SLIDE//05]
[TRACTION_2026]

NUMBERS DON'T LIE.

ZERO PAID ACQUISITION // ENTIRELY INBOUND + DEVELOPER COMMUNITY
87
Enterprise Customers
+52 in last 6 months
$48K
Avg Contract ACV
↑ from $28K (YoY)
14K
Slab GitHub Stars
+8K in 90 days
$1.8M
Pipeline — Signed
3 LOIs pending close
"We moved our entire 400-GPU training cluster to MONOLITH in 72 hours. Our monthly bill dropped 61%. Latency improved so dramatically we rewrote our SLA commitments to end users."
— VP Platform Engineering, Series-C AI Safety Co. [NDA]
[CUSTOMER_LOGOS // PARTIAL LIST]
INFERENCE-CO
FOUNDRY AI
ARKEN LABS
PELION HEALTH
VECTORA
AXIOM ROBOTICS
CREST FINANCIAL
+80 MORE
[SLIDE//06]
[COMPETITIVE_ANALYSIS]

WHY WE WIN
EVERY TIME.

DIRECT COMPARISON ON WHAT ACTUALLY MATTERS TO OPERATORS
CAPABILITY [MONOLITH] AWS SAGEMAKER GCP VERTEX AI AZURE ML COREWEAVE
P99 Latency — Inference 11ms ✓ 180ms 220ms 195ms ~80ms
Bare-metal GPU (no virt.) ✓ OWNED ✗ virtualized ✗ virtualized ✗ virtualized ✓ bare
Zero egress fees ✓ ZERO ✗ $0.09/GB ✗ $0.12/GB ✗ $0.08/GB partial
Gross margin (infra layer) 94% ~65% est. ~60% est. ~62% est. ~72% est.
SLA-backed latency credits ✓ AUTO ✗ ✗ ✗ ✗
No lock-in APIs ✓ OPEN ✗ proprietary ✗ proprietary ✗ proprietary partial
Cost vs. AWS equivalent BASELINE +310% +380% +290% +140%
[SOURCE: INTERNAL BENCHMARKS Q1 2026 // THIRD-PARTY AUDIT AVAILABLE IN DATA ROOM // HYPERSCALER PRICING FROM PUBLIC RATE CARDS]
[SLIDE//07]
[BUSINESS_MODEL]

94% GROSS MARGIN.
WE OWN THE HARDWARE.

[PRICING_TIERS]
DEV
Up to 8 GPUs / month-to-month
$2.20/GPU-hr
TEAM
Up to 64 GPUs / committed term
$1.85/GPU-hr
ENTERPRISE
Unlimited / custom SLA / dedicated cluster
Custom
SPOT
Preemptible / training bursts
$0.75/GPU-hr
Revenue mix: 68% Enterprise // 22% Team // 10% Dev+Spot
Enterprise avg ACV: $48K / yr
Gross margin expansion: +600bps YoY as HW depreciated
[UNIT_ECONOMICS]
94%
GROSS MARGIN — Q1 2026
Hardware is owned and fully depreciated over 5-year schedules. After year 3, incremental GPU-hours approach 98% margin.

Colocation contracts locked through 2031 at below-market rates. Power PPA signed at $0.042/kWh — 60% below US commercial average.

LTV/CAC = 18× at current NRR of 148%.
[COGS = POWER + COLO + NETWORK + SUPPORT]
[EXCLUDES: D&A, OPEX, HEADCOUNT]
[AUDITED BY KPMG — FULL REPORT IN DATA ROOM]
[SLIDE//08]
[MARKET_OPPORTUNITY]

THE LARGEST INFRASTRUCTURE
SHIFT IN A DECADE.

$420B
GLOBAL AI COMPUTE MARKET BY 2030 // CAGR 44% // IDC + GARTNER CONSENSUS
TAM
$420B
SAM
$62B
SOM (5yr)
$4.8B
SAM = bare-metal GPU cloud + dedicated AI infra.
SOM = 1.15% of SAM — achievable at $420M ARR by 2030
at current growth trajectory without further capital efficiency gains.
[COMPUTE SPEND BREAKDOWN 2026E]
AI TRAINING COMPUTE $28.4B
AI INFERENCE SERVING $21.3B
GPU-BACKED DATA PROCESSING $14.8B
FINE-TUNING + RLHF $7.9B
[TAILWIND] H100/H200 cluster availability from hyperscalers is >9 months backlog. MONOLITH has 2,400 GPUs available today with expansion capacity signed under LOI for 6,000 additional units by Q4 2026.
[SLIDE//09]
[FOUNDING_TEAM]

BUILT BY PEOPLE WHO
RAN THE INFRASTRUCTURE.

EK
ELENA KOSS
CO-FOUNDER & CEO
Former Director of Infrastructure at Stripe — oversaw $2.4B in annual compute spend. Negotiated Stripe's first colocation contracts at scale, cutting cloud costs 44% in 18 months.

Prior: Google SRE (Search infra, 5yr). Stanford MS CS. Two US patents in distributed scheduling.
MV
MARCO VIDAL
CO-FOUNDER & CTO
Creator of Slab — the open-source GPU scheduler with 14K stars adopted by xAI, Mistral, and Cohere for production inference. Former Principal Engineer at Meta AI (PyTorch infra team).

MS from ETH Zürich. Speaks at NeurIPS / SC annually.
PC
PRIYA CHANDA
CO-FOUNDER & CFO
Former VP Finance at Equinix — structured $4B in data-center financing across 3 continents. Architected colocation contract structures that became industry templates.

CPA, Wharton MBA. Ex-Goldman Sachs TMT banking. Closed MONOLITH's first 3 data center leases before we had customers.
[TEAM] 34 FTEs — 22 Eng // 6 Sales // 4 Ops // 2 G&A
[ADVISORS] Ex-CTO Rackspace // Former VP Eng Cloudflare // NVIDIA Academic Partner
[BACKERS] Conviction Capital (Seed) // Asymmetric (Seed+) // 12 angels
[SLIDE//10]
[ROADMAP_2026_2027]

24 MONTHS TO
CATEGORY DOMINANCE.

Q3 2026
[PLATFORM]
Launch Slab v2.0 — multi-tenant isolation + hot-swap GPU allocation
[EXPANSION]
DC #5 live — Frankfurt. 800 H100 units. EU data residency unlocked.
[MILESTONE]
$7M ARR target. 120 customers.
Q4 2026
[PRODUCT]
MONOLITH Managed Training — full fine-tuning pipeline on bare metal
[GTM]
AWS Marketplace listing live. 2,200 prospects in procurement pipeline.
[TEAM]
+12 engineers hired. Platform, infra, and ML systems.
Q1 2027
[COMPLIANCE]
FedRAMP Moderate authorized. Unlocks US federal + DoD pipeline.
[PRODUCT]
Dedicated private cluster offering — single-tenant, air-gapped. Target $250K+ ACV.
[MILESTONE]
$12M ARR target. Positive free cash flow.
Q2 2027
[RAISE]
Series B — $60M target at $300M+ valuation. 8× revenue multiple.
[SCALE]
8,000+ GPUs across 10 DCs. Asia-Pacific expansion (Tokyo, Singapore).
[MILESTONE]
250 enterprise customers. $18M ARR. Category defined.
[SLIDE//11]
[THE_ASK]
$18M
SERIES A  ·  $90M PRE-MONEY VALUATION  ·  20% EQUITY OFFERING
[USE_OF_FUNDS]
40%
Hardware Expansion
28%
Engineering Headcount
20%
Go-to-Market
12%
Ops / Legal / G&A
[24-MONTH_MILESTONES]
6,000 GPUs deployed across 8 DCs by Q1 2027
Adds $9M ARR capacity. Fully contracted colo sites.
$18M ARR by Q2 2027 (Series B readiness)
At current NRR of 148%, expansion covers >60% of target.
FedRAMP Moderate — Q1 2027
Unlocks $40M+ Federal pipeline already in qualification.
Positive FCF by Q3 2027
Hardware depreciation + expanding margins = self-funding inflection.
[TERMS_OVERVIEW]
$90M
PRE-MONEY VALUATION
21× trailing ARR ($4.2M)  ·  5× forward ARR ($18M target)

Lead commitment: $12M (Tier-1 fund, NDA).
Co-investor tranche: $6M — filling now.

Pro-rata rights for Series B reserved for Series A participants. Board seat offered to lead investor.
[CONTACT]
  investors@monolith.compute
  monolith.compute/deck
  Data room: secure.monolith.compute/sa-2026
  San Francisco · Austin · Amsterdam
[MONOLITH-SA-2026-V3.2 // CONFIDENTIAL // NOT FOR DISTRIBUTION]
[PAST PERFORMANCE DOES NOT GUARANTEE FUTURE RESULTS]
[FORWARD-LOOKING STATEMENTS SUBJECT TO MATERIAL RISKS]