LEAP Analytics GPU INTELLIGENCE TERMINAL
NVDA
AMD
B300 SPOT AVG
$7.00/hr
rental market
B300 ASP
$50K
purchase price
H100 SPOT / LTC
$2.15
LTC $2.40 ↑
SECONDARY MARKET VALUE
RENTAL RATES
FACTOR MODEL
EQUITIES
FINANCING
DEAL FLOW ⬡
METHODOLOGY
SECONDARY MARKET RESALE VALUE — $/UNIT
SHOW: LEAPEF QUANT MODEL
GPUPurchase PriceToday+6 Mo+12 Mo+18 Mo+24 Mo+36 Mo+48 MoFloorFloor %
B300 SECONDARY PRICE PROJECTION — DOTTED = LEAPEF MODEL FROM APRIL 2026
Projection starts today. Dot = current purchase price. Cone = P25–P75 confidence band.
Model factors: Generation depreciation curve (calibrated from A100/H100 observed data) · CoWoS packaging capacity · HBM memory ASP · ASML EUV shipments · Helium/Neon supply · Hyperscaler capex · Next-gen performance delta · LEAPEF Deal Flow Index (if unlocked). Adjust all factors in the Factor Model tab.
GPU CLOUD RENTAL RATES — EX-AWS · 55 PROVIDERS · MAR 2026
GETDEPLOYING.COM
GPUArchitectureVRAMSpot MinSpot AvgSpot Max†LTC RateProvidersBuild CostAvg/BuildStage
† Excludes outlier hyperscaler pricing. LTC = Long-Term Contract. Build Cost = 5yr TCO/hr.
H100 MARKET STRUCTURE — SPOT vs LTC vs COST-TO-BUILD
UPPER90 / SEMIANALYSIS · KEY INVESTOR REFERENCE
RENTAL RATE FORWARD CURVE — $/GPU-HR · PROJECTION FROM APRIL 2026
Dot = today's rate. Dotted line = LEAPEF model projection.
SUPPLY & DEMAND FACTOR MODEL — ADJUST SLIDERS · CHARTS UPDATE
AFFECTS B300 FLOOR PRICE
SUPPLY FACTORS
CoWoS Packaging Capacity (TSMC)
CRITICAL · 97%
Weight β=+0.42 · bottleneck #1
Loose 40%Critical 100%
HBM Memory ASP (SK Hynix / Micron)
TIGHT · $22/GB
Weight β=+0.38 · bottleneck #2
$8 loose$35 tight
ASML EUV Shipments (machines/yr)
CONSTRAINED · 60/yr
Weight β=−0.18 · more = more supply
30/yr120/yr
Helium Spot Price $/mcf (Qatar risk)
WATCH · $280/mcf
Weight β=+0.07 · Hormuz tail risk
$150 normal$700 shock
Neon Gas Spot $/m³
RECOVERING · $14/m³
Weight β=+0.05 · post-Ukraine
$8 normal$120 shock
NVIDIA Units Shipped K/quarter
RAMPING · 650K/Q
Weight β=−0.35 · more = lower price
100K tight2,000K flood
DEMAND FACTORS
Hyperscaler Capex $B/yr combined
STRONG · $320B
Weight β=+0.32 · MSFT+GOOG+META+AMZN+ORCL
$100B slow$600B surge
Next-Gen Performance Delta
MODERATE · 2.0×
Weight β=−0.25 · bigger leap = more pressure
1.0× same5.0× leap
Inference Demand Growth %/yr
STRONG · 85%/yr
Weight β=+0.28 · decimalization effect
10%/yr200%/yr
Supply Pressure Index
Higher = tighter supply = higher price
Demand Pull Index
Higher = stronger demand = higher price
Composite Score
Net factor impact
B300 Floor Adjustment
vs base floor $21,000
FACTOR SENSITIVITY — $/UNIT IMPACT ON B300 SECONDARY FLOOR
NVDA Price
Loading…
NVDA 52W High
AMD Price
Loading…
AMD 52W High
NVDA/AMD Ratio
Price multiple
Monte Carlo
500
paths · 12M horizon
NVIDIA (NVDA) — 1YR DAILY + MA50/200 + MONTE CARLO 12M FORWARD
LIVE · ALPHA VANTAGE
Fetching NVDA from Alpha Vantage…
AMD — 1YR DAILY + MA50/200 + MONTE CARLO 12M FORWARD
LIVE · ALPHA VANTAGE
Fetching AMD from Alpha Vantage…
NVDA vs AMD — NORMALISED RELATIVE PERFORMANCE (START=100)
LIVE
FINANCING PARAMETERS
FINANCING OUTPUT
COMPUTED
Set parameters and click CALCULATE
LEAPEF PROPRIETARY DEAL FLOW
Confidential contract data influencing the forward price model.
Session-only. Never stored, transmitted, or logged.
1. The Core Question
"I buy 1,000 B300 GPUs at $50,000 each. What can I sell them for in 12, 24, and 36 months?"

This is the primary investor concern in GPU-backed structured finance. The tool answers it through a multi-factor model combining historical GPU depreciation curves with supply chain and demand inputs.
2. Secondary Market Price Model
SECONDARY_PRICE(t) = (P₀ − F_eff) × e^(−λ_eff · t) + F_eff

P₀ = Purchase price today · F_eff = Effective floor (factor-adjusted) · λ_eff = Decay rate (factor-adjusted) · t = Months from purchase
Base parameters calibrated from: A100 (4+ years observed, floor confirmed at ~$3,800) and H100 (27 months observed, floor forming at ~$14,000–18,000). These two calibration points validate the model before applying forward to B300.
3. The Factor Model
F_eff = F_base × (1 + Σ βᵢ · Δfactorᵢ) · λ_eff = λ_base × (1 − Σ γⱼ · Δfactorⱼ)

Supply: CoWoS β=+0.42 · HBM β=+0.38 · ASML β=−0.18 · Helium β=+0.07 · Neon β=+0.05 · Units shipped β=−0.35
Demand: Hyperscaler capex γ=+0.32 · Perf delta γ=−0.25 · Inference demand γ=+0.28
4. Monte Carlo — The Confidence Bands
P(t+1) = max[(P₀−F_eff)·e^(−λ_eff·t) + F_eff + σ·√(1/12)·Z, F_eff·0.80] · Z~N(0,1)
1,000 paths · P10/P25/P50/P75/P90 bands
P50 = central estimate for base-case investor presentations. P25 = cautious case for covenant floor analysis. The cone width represents genuine uncertainty — it grows over time because predicting markets 36 months out is inherently uncertain. LEAPEF does not present false precision.
5. Why GPUs Do Not Go to Zero
The asymptotic floor prevents projection below economic viability. Workload stratification: Each new generation pushes prior generations into larger, more price-elastic workload markets — not obsolescence. H100 today serves the role A100 served in 2023. B300 will serve the role H100 serves today. A100 validation: Still trading at $3,800–4,500 despite being 4+ years old and 3 generations behind the frontier.
6. LEAPEF Deal Flow Index
DFI = Σᵢ[wᵢ×(contract_priceᵢ/market_priceᵢ)] / Σwᵢ · wᵢ = units×tenor×confidence
DFI>1.0 → deal flow above market (bullish) · Floor adjustment: (DFI−1.0)×$3,000/unit
PIN-gated. Session-only. Never stored or transmitted. The DFI value is the only publicly visible output.
Disclaimer

LEAPEF AB internal tool only. Does not constitute investment advice, a credit rating, or a formal valuation. All projections are probabilistic estimates. LEAPEF AB is a registered AIFM (Finansinspektionen). March 2026.