Deeptech product portfolio · Quantum × Blockchain × AI

v1.0 · August 2026

Sixteen products that are 80% software now and 20% hardware later.

Most deeptech pitches fail one of two ways: pure software with no defensible edge, or hardware-first with nothing to show until the capital lands. This portfolio takes a third path — products whose hardware dependency is real but deferrable. Every one has a named software surrogate for its hardware, so it demos, sells, and wins reference customers before the capex arrives.

16Use cases
4Shared primitives
5Need no hardware
9Dated obligations
7 wkFastest demo

The thesis

Why the deferrable-hardware path works in 2026

  1. The quantum money is in the migration, not the machine. No cryptographically-relevant quantum computer exists or is close. But India's DST Task Force (Feb 2026) has put dates on the wall — critical-infrastructure inventory by 2027, high-priority migration by 2028, full PQC by 2029 — with CBOM submissions from vendors from FY 2027-28. NSA CNSA 2.0 requires networking equipment by 2030 and compliance in new national-security-system acquisitions from 1 Jan 2027. That is budget with a date on it, and it is entirely software work.
  2. The blockchain money is in regulatory non-repudiation, not tokens. The EU Battery Passport is adopted law with an 18 Feb 2027 deadline. DSCSA interoperable EPCIS is already mandatory. EUDI Wallet obligations land Dec 2026 with fewer than a third of member states ready. None of these regulations mandate a ledger — which is exactly why the ledger should be a thin attestation layer inside a conventional data platform, and why the winner is whoever ships the data platform.
  3. The AI money is in the control plane, not the model. Agentic operations are shipping but unsafe: AI-attributed incidents rose from 1.7% of disclosed incidents in 2023 to 10.7% in 2026 YTD, with at least nine documented cases of autonomous agents destroying production data — nearly always by holding credentials they should never have had. The missing layer is authorization, blast-radius computation, and cryptographic non-repudiation of agent actions.

Convergence here is not decoration. In every use case below, at least two of the three technologies are load-bearing: remove one and the product stops working, or stops being defensible.

The operating rule

Rate the 20% before you pitch it

The hardware tier is the single most important number in a deeptech pitch, because it tells an investor exactly what their money buys. Every card below carries one.

HardwareSurrogate for the 80% phaseWhat you can honestly claim
QPUGPU annealer, simulated bifurcation, cuOpt, tensor-network simulatorSolution quality and time-to-solution today; the QPU is a pluggable backend behind the same API
QRNGNIST SP 800-90B DRBG behind an entropy-source abstractionCorrect architecture and a swappable source — never that the entropy is quantum
TEE / confidential GPUAttestation stub carrying the full chain-of-trust structure, unsignedThe protocol is correct; the attestation is explicitly marked unverified
Field sensor fleetReplayed public datasets plus a physics-based synthetic trace generatorDetection performance on synthetic and historical data, stated as such
Fab / ATMP telemetryPublic wafer-map datasets plus synthetic die genealogyAlgorithmic performance — not yield economics
Quantum sensorClassical sensor at a stated lower signal-to-noise ratioThe pipeline works; the sensitivity upgrade is the investment thesis

The substitution principle: every hardware dependency needs a named software surrogate that is honest about what it does and does not prove. The discipline is the deliverable — a demo that says "this runs on a GPU annealer today and a QPU when the economics invert, here is the cost model for both" earns more trust than one implying a quantum computer is in the loop.

Platform economics

Four primitives, sixteen products

This is not sixteen projects. It is four reusable primitives and sixteen vertical applications assembled from them — build once, sell many times.

Reuse is the thesis. P3 carries twelve of sixteen products; hardening it once pays for itself four times over.

The portfolio

Sixteen converged use cases

Grouped into three clusters. Filter by vertical, by which technologies are load-bearing, or by how much hardware capital they need.

Cluster
Stack
Hardware

No use case matches that combination. Clear a filter to see more.

Decision tool

Rank and sequence them against your own constraints

Set what you actually optimise for and everything below recomputes — the ranking, what each candidate scores on, and a build schedule that respects your team size and capital budget.

What you optimise for
Your constraints

Projects are crewed at three engineers, so team size sets how many run at once. Capital is charged in ranked order — anything past the budget line is drawn hatched and gated on the next raise.

Bars are normalised 0–100 across the sixteen candidates, so they compare positions within this portfolio — not against the wider market.

Why now

The deadlines are the demand signal

Nine of sixteen products sit behind a dated regulatory or standards obligation. None of it is discretionary budget, and none of these dates move for a vendor who is not ready.

Portfolio construction

Build order, so each wave funds the next

Sixteen products is a map, not a plan. Sequence so zero-hardware revenue arrives first, deadline-driven products follow on seed money, and the capital-intensive work is pitched only once reference customers exist.

The anti-portfolio

What not to build — and why saying so wins the room

Naming these publicly is worth more than any single product, because it makes every other claim on this page credible.

Communication

Six talks that fall out of this portfolio

Verification

Sources

Every date and figure on this page traces to a primary or near-primary source, checked in August 2026.