Phase A — Understand the business
Lens 1 · Company Overview
LanzaTech is a gas-fermentation carbon-recycling company. Its core technology uses proprietary engineered bacteria to ferment carbon-rich waste gases — steel-mill and refinery off-gas, syngas from gasified waste, captured CO/CO2 — into ethanol and other chemical intermediates, which become feedstock for acrylics, plastics, synthetic rubber, packaging, and (via its LanzaJet affiliate) sustainable aviation fuel (SAF). The pitch: let an industrial emitter bolt a LanzaTech unit onto an existing smokestack and turn a waste/cost stream into a revenue stream — "a single process producing multiple chemicals from diverse feedstocks, stable despite fluctuating gas compositions, unlike thermocatalytic methods".
Founded 2005 in New Zealand, now HQ Skokie, Illinois; went public Feb 2023 via SPAC merger with AMCI Acquisition Corp. II. Delaware incorporated, Nasdaq: LNZA (+ warrants LNZAW).
Revenue model (four lines, FY2025):
- Contracts with customers & grants — $18.3M (engineering/licensing services + government grants; down from $23.0M FY24).
- CarbonSmart product sales — $14.6M (resale of ethanol/derivatives produced at partner plants; nearly doubled from $7.9M).
- Collaborative arrangements — $2.4M (co-development; down from $5.6M).
- Related-party transactions — $20.5M (37% of total; up from $13.1M — flagged in Lens 9/10).
Customers / partners named in the filing: ArcelorMittal, BASF, IndianOil, Mitsui (alliance partner in Japan), Sekisui (Japan), plus a first commercial facility in China that has sold >63 million gallons of ethanol to date — the single best proof point that the chemistry works at commercial scale. Strategic financial partner: Brookfield (framework agreement to fund commercial facilities; also a lender — see Lens 5).
Contract structure: the new strategic spine is a "cohort-based commercialization model" — licensing the platform to groups of industrial partners rather than building/owning plants itself (asset-light, but it makes revenue dependent on partners' execution and timing). Government-grant revenue carries shutdown/funding-cycle risk the company explicitly calls out.
This is a single reportable segment — the CODM (CEO) reviews performance at the consolidated level.
Lens 2 · Supply Chain
Map (upstream → LanzaTech → end customer), with named stakeholders:
Upstream inputs:
- Waste-gas feedstock suppliers = the industrial emitters themselves: steel mills (ArcelorMittal; the China JV plant), refineries, gasified-waste sites in China, India, Belgium. Feedstock is effectively free/negative-cost (it's pollution), which is the moat's economic root — but single-source per site: each plant is captive to one host emitter's gas stream and uptime.
- Microbe / biocatalyst = LanzaTech's own engineered strains + proprietary media (in-house; the trade-secret crown jewel).
- Bioreactor hardware / equipment packages = engineered and sold by LanzaTech; fabrication outsourced (CDMO-style equipment vendors, unnamed in filing).
The company: licenses the process + supplies the microbe, media, engineering and equipment package; books licensing/engineering revenue up front and a CarbonSmart product margin on output it markets.
Downstream:
- Plant operators / licensees = the host emitters and JV entities who run the units (China JV, India, Belgium). LanzaTech "relies heavily on industry partners to effect our growth strategy" — explicitly its largest execution dependency.
- Chemical/material off-takers = buyers of ethanol and CarbonSmart derivatives (apparel/packaging brands historically; e.g., the Sekisui collaboration).
- LanzaJet (37.5%-owned affiliate) = the alcohol-to-jet SAF route; LanzaTech contributed IP for its stake.
- Mitsui = Japan go-to-market alliance; LanzaTech may not recommend an alternative provider of the covered services without Mitsui's written consent (a channel lock that cuts both ways).
Chokepoints / single-source dependencies:
- Partner plant uptime — revenue recognition keys off partners commissioning and running facilities; LanzaTech does not control the asset.
- The microbe + media — concentrated in LanzaTech's own trade secrets (a moat, but also a single point of technical failure / key-person risk).
- Government incentive regimes — the unit economics of low-carbon fuels lean on subsidies (CCU/SAF credits); a policy reversal hits demand directly.
Names are present, so this lens passes — but the chain is thin and partner-dependent, which is the structural fragility of the whole story.
Lens 3 · Competitive Advantages (moats)
What's genuinely defensible:
- IP estate — 616 granted patents + 190 pending (owned & in-licensed), spanning 118 patent families across US/EU/Asia, plus trade-secret microbe/media know-how. For a platform-biology company this is the real asset; arguably the IP + the strains are worth more than the equity.
- Proven scale — >100,000 hours of pilot/demo operation and a commercial plant that has shipped >63M gallons of ethanol. Very few synbio names ever cross from lab to commercial tonnage; LanzaTech has. This is its single strongest differentiator versus the synbio graveyard (Amyris, Zymergen, Solazyme all died at the scale-up step).
- Feedstock flexibility — one organism, many gas compositions; management argues no competitor delivers feedstock + product + synbio + process + commercial scale "in an integrated way".
- Negative-cost feedstock — waste gas is free, structurally undercutting fossil and crop-based routes when carbon is priced.
Where the moat is weak:
- Bargaining power is inverted. LanzaTech needs the emitters more than they need it — partners control the plants, the gas, and the timing. The Mitsui consent lock and Brookfield framework show a company negotiating from weakness, trading exclusivity for capital/channel.
- No pricing power as a microcap in distress — counterparties know the balance sheet (going concern is public) and can extract terms.
- Commodity output — ethanol is a price-taker; the margin lives in the carbon credit / subsidy, not the molecule.
Moat verdict: a genuine technology/IP moat wrapped around a commercially weak position. The asset is real; the ability to capture value from it is not yet proven.
Lens 4 · Segments
Single reportable segment (consolidated), so the meaningful breakout is by revenue line and the trend:
| Revenue line | FY2025 | FY2024 | YoY | Trend / cause |
|---|
| Contracts w/ customers & grants | $18.3M | $23.0M | −20% | Decelerating — weaker licensing/engineering activity |
| CarbonSmart product sales | $14.6M | $7.9M | +84% | Accelerating — the one bright line; product resale scaling |
| Collaborative arrangements | $2.4M | $5.6M | −56% | Shrinking — fewer co-development dollars |
| Related-party transactions | $20.5M | $13.1M | +57% | Growing — but it's affiliate/VIE revenue (quality flag) |
| Total revenue | $55.8M | $49.6M | +13% | Top line up, but mix deteriorating |
[All rows: research-layer: filings/10-k-2025-q4.md, Consolidated Statements of Operations]
The uncomfortable read: headline growth (+13%) is flattered by a +57% jump in related-party revenue. Strip that out and arm's-length revenue went $36.5M → $35.3M, i.e. roughly flat-to-down. The genuinely-external growth engine (CarbonSmart, +84%) is real but small ($14.6M). Q1-2026 confirms the wobble: contracts/grants jumped to $7.3M (from $3.1M) on grant timing, but related-party fell to $0.7M — lumpy, project-driven, not yet a durable run-rate.
Geography is not separately reported, but operations are explicitly China / India / Belgium (plants) + Japan (Mitsui/Sekisui) + US (HQ/grants).
Phase B — Measure performance
Lens 5 · Earnings Result (latest print = Q1 2026, filed 2026-05-14)
[All figures: research-layer: filings/10-q-2026-q1.md, Consolidated Statements of Operations & Balance Sheet]
- Revenue $12.0M vs $9.5M Q1-2025 (+27%) — driven by contracts/grants ($7.3M vs $3.1M); CarbonSmart roughly flat ($4.1M); related-party collapsed to $0.7M.
- Loss from operations $(9.8)M vs $(31.1)M — a 68% narrowing, the whole story of the quarter.
- Net loss $(14.7)M, EPS $(1.77) vs $(9.79); the gap between op-loss and net-loss is the LanzaJet equity-method drag $(4.5)M + small other-expense.
- Cost takeout is dramatic: R&D $4.0M vs $16.5M (−76%); SG&A $8.6M vs $15.7M (−45%). Total opex $21.8M vs $40.5M. Management/press frame this as a ~59% opex reduction from the 2025 restructuring.
- Balance sheet: cash $19.9M (up from $13.2M at YE25, thanks to the Jan-2026 PIPE); trade receivables $10.6M; accumulated deficit $(1.033B).
- Operating cash burn Q1: $(9.3)M → at that pace, $19.9M is ~2 quarters of runway before the next raise; with the May-2026 $10M top-up (Lens 9), ~3 quarters.
- Guidance / tone: "streamlining business priorities, reducing cost structure, evaluating liquidity-enhancing initiatives including capital raising, partnership or asset-related opportunities, and other strategic options" — i.e., explicit strategic-review / possible-asset-sale language.
- Going concern: NOT alleviated. The FY2025 10-K is blunt: "existing cash will not be sufficient to fund operations through the next twelve months… these conditions raise substantial doubt". The Q1-2026 Q repeats it.
- Market reaction / context: the cost-cut narrative + repeated raises have stabilized the post-split price (~$5–7), but the 52-week tape is −85%.
The honest summary: the loss is shrinking because the company is shrinking (R&D down 76%), not because the business is inflecting. That is a survival result, not a growth result.
Lens 6 · Earnings Calls (sentiment trend)
No transcripts on disk (transcripts/ empty). From web grounding of the call/release cadence:
- Tone shift (2023 → 2026): from SPAC-era growth/TAM evangelism ("circular carbon economy," $86.5M FY24 revenue hopes) → to a defensive survival framing: cost discipline, runway, "strategic options," asset-light cohort licensing. The vocabulary that disappeared: aggressive revenue-ramp guidance, new-plant announcements as cadence. The vocabulary that arrived: "going concern," "restructuring," "59% opex reduction," "evaluating strategic options."
- What management is focused on now: (1) not running out of cash; (2) the cohort-licensing pivot as the path to operating leverage; (3) keeping the Nasdaq listing; (4) monetizing the LanzaJet/IP optionality.
- Consensus has capitulated alongside: 4 analysts no longer model breakeven; 2026 loss consensus ~$60.2M.
``-labeled throughout; absent transcripts, treat sentiment as directional, not quote-exact.
Lens 7 · Comps
No clean public-comp multiples exist for a going-concern microcap, and LanzaTech is loss-making at every line, so EV/EBIT and P/E are n/a (negative). The honest comp set is the industrial-synbio / carbon-management cohort — by mechanism and fate, not by P/E:
| Company | Ticker | ~Mkt cap | EV/Sales | P/E | Note |
|---|
| LanzaTech | LNZA | ~$77.5M | ~1.4× on $55.8M rev | n/a — loss-making | Going concern; cohort-licensing pivot |
| Gevo | GEVO | n/a | n/a | n/a | FCF-positive Q4-25; ~$30M adj-EBITDA target 2026 |
| Aemetis | AMTX | ~$200M | n/a | n/a | Biofuels/RNG/CCS; catalyst-driven |
| Amyris | (delisted) | — | — | — | Chapter 11, Aug 2023 — burned $3B+ on precision-fermentation scale-up |
| Zymergen / Solazyme | (defunct) | — | — | — | Synbio cautionary tales — died at commercialization |
[Multiples I cannot source are marked n/a rather than fabricated.]
The comp table's real message isn't a multiple — it's survivorship. The reference class for "platform synbio that raised big and tried to scale" is littered with bankruptcies (Amyris, Zymergen). LanzaTech's relative claim to live is that it actually reached commercial tonnage and has an asset-light pivot; Gevo's counter-example shows the sector can reach FCF-positive, which is the bull's existence proof. At ~$77.5M market cap on $55.8M revenue (~1.4× EV/Sales) the equity is priced as an option, not a going concern — cheap on sales, but sales quality and solvency are the issue, not the multiple.
Lens 8 · Stock-Price Catalysts (the >5% movers)
- Feb 2023 — SPAC debut, −21% day one. The market rejected the de-SPAC valuation immediately.
- 2023–2024 — serial guide-downs: FY24 revenue consensus cut $86.5M → $62.9M; the stock bled with each downward revision.
- Mar 2025 — Nasdaq sub-$1 bid-price deficiency notice → overhang.
- Aug 18 2025 — 1-for-100 reverse split to cure the bid price; regained compliance Sep 3 2025. Reverse splits are usually a downward catalyst on the underlying trend even when the print "works."
- Jan 2026 — $20M PIPE @ $5.00 + preferred conversion (dilution/concentration).
- Q1-2026 print (May 14 2026) — 59% opex cut, loss narrowed — the stabilization catalyst; price held ~$5–7.
- Auditor change (Deloitte → BDO) for FY2026 — a governance/going-concern signal the market watches.
Pattern: for LNZA the tape reacts to solvency and dilution events first, fundamentals second. Earnings beats on a shrinking base don't re-rate it; what moves it is "will it survive and how much will I be diluted." That tells you the equity is trading as a distressed-financing option, and the catalysts that matter are capital events, a strategic transaction, or a delisting — not a revenue beat.
Phase C — Judge people & books
Lens 9 · Management
- CEO & Board Chair: Dr. Jennifer Holmgren. The long-time face of LanzaTech and Board Chair of LanzaJet — deep scientific/sustainability credibility (a genuine domain authority who took the platform from lab to the first commercial plant). Track record is real on the science, unproven on shareholder value: she built a technology that works at scale, but the public equity is down ~95%+ from the de-SPAC and the company is at going concern. Concentrating Chair+CEO+affiliate-Chair in one person is a governance concentration to note.
- CFO: Sushmita Koyanagi (effective June 2025); interim General Counsel: Amanda Fuisz (June 2025). A finance/legal refresh consistent with a company in restructuring/strategic-review mode.
- Board: Gary Rieschel retired post-2025 AGM; Reyad Fezzani (Regenerate Power) added Jan 2025.
- Skin in the game / control: Khosla Ventures and affiliates have significant influence/control — explicitly flagged as a risk factor (their interests may conflict with other holders). Post Jan-2026 PIPE + preferred conversion, ownership is highly concentrated in insiders/strategics (Khosla, "LT Global SPV") — the float is tiny (~10M shares). This is the double-edge of the whole thesis: insiders are the lifeline and the controllers.
- Capital-allocation history: value-destructive at the corporate level. Accumulated deficit $(1.033B); burned the entire $242M de-SPAC trust plus subsequent raises; LanzaJet IP contribution was strategically sound but hasn't yet returned cash. The 2025 restructuring (−59% opex) is the first genuinely disciplined capital-allocation move and is to their credit.
- Founder vs professional: founder/scientist archetype (Holmgren) — strong on technology vision and partnerships, historically weak on capital discipline and commercial monetization. For this stage (survival → asset-light pivot), the recent CFO-led cost discipline matters more than vision.
Lens 10 · Forensic Red Flags
The red flags are unusually concentrated and material:
- Going concern — explicit, twice. "Existing cash will not be sufficient to fund operations through the next twelve months… substantial doubt". This is the headline forensic fact.
- Material weaknesses in internal control over financial reporting — explicitly disclosed; remediation pending. ICFR weakness + a microcap in distress is exactly the combination forensic analysts weight most.
- Auditor change mid-stream — dismissed Deloitte & Touche, engaged BDO USA for FY2026; Deloitte's FY24/FY25 opinions carried the going-concern explanatory paragraph. Auditor turnover during a going-concern/ICFR period is a yellow-to-amber flag.
- Related-party revenue = 37% of FY2025 sales ($20.5M). Revenue from affiliates/VIEs (the LanzaJet/JV orbit) is lower-quality than arm's-length; it props the top line and grew +57% while external revenue was flat. This is the single most important accounting-quality flag — the growth narrative partly depends on transactions with entities the company itself influences.
- Earnings quality / non-cash flatter: FY2025 net loss $(49.0)M vs loss-from-operations $(79.2)M — the $30M "improvement" is $41.5M of "other income" (warrant/derivative/FPA fair-value swings), non-operating and non-cash. Cash burn ($(64.9)M operating) tells the true story; the GAAP net-loss narrowing overstates the turn.
- Liability structure is exotic and dilution-laden: FPA Put Option liability $30.0M, Brookfield Loan $10.9M (replaced a terminated SAFE), convertible preferred ($13.2M mezzanine), PIPE warrants. Multiple instruments that convert to equity = a dilution overhang baked into the cap structure.
- Serial dilution at rising marks: Jan-2026 $5.00, May-2026 $10.00, plus a registered-direct follow-on — repeated issuance is the funding mechanism, and each round concentrates control further.
Regulatory findings (required sub-section):
- SEC Litigation Releases / AAERs: None.
regulatory/regulatory-findings.md (EDGAR EFTS, LR + AAER, period 2021-06-20→2026-06-20) returns 0 SEC findings.
- Item 3 Legal Proceedings (10-K): no material litigation — "normal course of business… do not believe the outcome will have a material adverse impact," incorporated to Note 17.
- Non-SEC enforcement (web): no material FTC/DOJ/FDA/EPA enforcement action surfaced for LanzaTech in search.
- Conclusion: No material regulatory or legal-enforcement findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 as of 2026-06-20. The forensic risk here is solvency, ICFR weakness, related-party revenue, and dilution — not fraud or enforcement.
Phase D — Project & stress-test
Lens 11 · Forward Projection (FY2026–FY2028)
Bottom-up from Q1-2026 actuals + the restructuring run-rate. No forecast.ts logged (watchlist/breadth mode — per skill rules, the Brier forecast is only created on a genuinely committed base case; this is a survival-binary, not a clean EPS line).
Inputs (all ``, arithmetic shown):
- FY2026 revenue: Q1 $12.0M; lumpy grant timing makes annualizing unreliable. Consensus 2026 loss ~$60.2M. Revenue base case ~$48–55M (flat-to-down vs FY25 as related-party normalizes).
- Opex at restructured run-rate: Q1 opex $21.8M → ~$85–90M annualized, but management is still cutting; assume ~$80M FY26.
- Operating loss FY26 ~$(30)–(40)M; cash burn ~$(35)–(45)M before financing.
| Scenario | FY2026 | FY2027 | FY2028 | Driver |
|---|
| Bull | op-loss ~$(25)M | ~$(10)M | ~breakeven | Cohort licensing scales; CarbonSmart + grants ramp; cost base holds; a strategic deal/asset sale injects cash. EPS still negative but FCF approaches zero. |
| Base | op-loss ~$(35)M | ~$(28)M | ~$(22)M | Revenue flat ~$50M, related-party fades, opex grinds down slowly; perpetual dilutive financing; survives but doesn't inflect. |
| Bear | op-loss ~$(40)M | n/a — capital event | n/a | Runway runs out before cohort traction; forced asset sale / recap / delisting; equity impaired toward zero. |
The number that actually matters is not EPS — it's runway-to-self-sustaining. Cash $19.9M (Q1) + $10M (May LT Global) + up-to-$20M LT Global option (conditioned on low cash) ≈ $30–50M of accessible liquidity against ~$35–45M/yr burn = roughly 12–15 months unless the burn keeps falling or a strategic deal lands. EPS base case FY2026 ≈ $(4)–(6) on ~12M shares, but dilution makes per-share figures nearly meaningless.
Falsifiable base call: LanzaTech does NOT reach operating-cash-flow breakeven by FY2028 and requires at least one further dilutive raise or a strategic transaction before then. (Not logged to forecast.ts per watchlist rules.)
Lens 12 · Bull vs Bear
Bull case (narrative). This is the only gas-fermentation platform that crossed the chasm synbio kills everyone at — >63M gallons shipped, 100k+ operating hours, 600+ patents, blue-chip partners (ArcelorMittal, BASF, Mitsui, IndianOil). The 2025 restructuring proved management can cut to a real cost base (−59% opex), and the asset-light cohort-licensing model — sell the platform, don't build the plants — is the high-operating-leverage path that, if even two or three cohorts close, flips the P&L. Insiders (Khosla, Brookfield, LT Global SPV) keep funding it because the IP + LanzaJet SAF optionality is worth a multiple of the $77M cap in a carbon-priced, SAF-mandated world. At ~1.4× sales with a tiny float, any sign of cohort traction or a strategic/IP transaction is a multi-bagger. Gevo's swing to FCF-positive is the sector existence-proof that this can work.
Bear case (narrative). Strip the related-party revenue and the business is flat-to-shrinking; the "improvement" in net loss is non-cash fair-value gains, not operations. It is a going concern with material-weakness ICFR and a fresh auditor change, kept alive only by serial dilution from the very insiders who control it — minority holders are along for a value-destructive ride (accumulated deficit $(1.03B)). The cohort model is a promise, not a run-rate; revenue is lumpy, grant-dependent, and partner-controlled. The reference class (Amyris, Zymergen, Solazyme) is bankruptcy. The float is so concentrated that price is technical, not fundamental.
Pre-mortem (18 months out, thesis broke): Cohort licensing didn't convert to bookings; grant revenue slipped on a funding cycle; cash hit the LT Global trigger and the company did a deeply dilutive recap or sold the IP/LanzaJet stake at a markdown; Nasdaq compliance lapsed again; the equity is a sub-$2 stub.
Are multiples too high? On sales, no (~1.4×). On quality of those sales and solvency, the equity is still expensive relative to the probability-weighted outcomes — it's an option priced as if survival is likely.
Contrarian view (what the market refuses to see): the IP + LanzaJet stake may be worth more than the whole equity in a strategic/IP sale — the bear-case "forced asset sale" could actually be the bull catalyst for a patient holder, because a trade buyer (a major, a chemicals/energy strategic) gets a de-risked platform for a distressed price. The asymmetry isn't "growth re-rate," it's "M&A/IP-monetization at a premium to a $77M cap."
Lens 13 · Devil's Advocate (short-seller)
Dismantling the bull case:
- Revenue quality is the kill shot. 37% related-party, growing +57% while external revenue is flat — the top line is partly a related-party construction. A short argues the "growth" evaporates the moment affiliates stop transacting (Q1-2026 already showed related-party collapsing to $0.7M).
- The cohort model has no proof of bookings. It's the third strategic framing in three years (build → collaborate → license). Each pivot is a tell that the prior model didn't generate cash.
- Dilution is structural and bottomless. FPA put ($30M), Brookfield loan, convertible preferred, PIPE warrants, the LT Global $20M option — every financing is equity-linked and at the insiders' discretion. Minority holders cannot win a coin-flip where the house keeps re-dealing.
- ICFR material weakness + auditor change during going concern is the classic late-stage distress signature.
- Most dangerous competitor bulls underestimate: not a synbio rival — it's the do-nothing alternative + cheap fossil ethanol. When carbon isn't priced high enough or subsidies wobble, the host emitter's payback math breaks and the whole licensing pitch stalls.
- Price-disappoints-20–30% math: the equity is already −85% YoY; another 20–30% on a missed raise or a delisting and it's a sub-$4 stub heading to the FPA-put/recap scenario.
- The single permanent-impairment scenario: cash hits the LT Global trigger before cohort revenue materializes → forced recap or IP sale at distress price → existing equity wiped/heavily diluted. Plausibility: moderate-to-high over 18–24 months absent a strategic deal.
The short's problem (and why I'm WATCHING not outright BEARISH): the float is tiny and insiders keep funding — a low-cost-to-borrow short can be squeezed on any good-news pop, and the IP-sale optionality caps the downside-to-zero thesis.
Lens 14 · Management Questions (ordered by information value)
- What is the contracted, signed-bookings backlog of the cohort-licensing model today (not pipeline) — in dollars and number of partners — and when does it convert to cash revenue?
- Excluding related-party transactions, what was arm's-length revenue in FY2025 and Q1-2026, and what's the FY2026 arm's-length target?
- At the current burn, on what date does cash hit the LT Global $20M-option trigger, and what's the dilution if you draw it in full?
- Is the board running a formal strategic review / sale process for the company, the IP estate, or the LanzaJet stake — and what bankers/advisors are engaged?
- What is the specific remediation plan and timeline for the material weaknesses in ICFR, and what drove the Deloitte → BDO auditor change?
- What is the realizable value of the LanzaJet 37.5% stake, and under what conditions would you monetize it?
- What is the path to operating-cash-flow breakeven — revenue level, gross-margin assumption, and date — and what has to be true for it?
- How much of the related-party revenue is recurring vs one-time, and who are the counterparties?
- What are the CarbonSmart unit economics (gross margin), and can that line scale to a meaningful fraction of revenue?
- What governance guardrails protect minority holders given Khosla/LT Global control and the equity-linked financing structure?
- Which government incentive programs is revenue most exposed to, and what's the downside if SAF/CCU credits are cut?
- What is the gross-margin and uptime track record of the operating partner plants (China/India/Belgium), and how dependent is recognition on their performance?
- How defensible is the microbe/media trade-secret moat against well-capitalized entrants, and what's the patent-cliff timeline on the core families?
- What's the realistic timeline and capex for the next cohort commercial facility to reach steady-state?
- If you could only fund one of {cohort commercialization, CarbonSmart scale-up, LanzaJet/SAF, core R&D} for the next 18 months, which — and why?