Meta AI / FAIR

Phase A — Understand the business

Lens 1 · Company Overview

Meta's AI franchise sits inside a business that, at the consolidated level, generates substantially all revenue from advertising. FY2025 total revenue was $200.97B (advertising $196.18B, other revenue $2.58B, Reality Labs $2.21B). The AI unit is therefore not a P&L line — it is an embedded capability with two faces:

• The cost side (FAIR + Meta Superintelligence Labs, MSL): Fundamental AI Research (FAIR) is Meta's long-horizon lab; in June 2025 Zuckerberg created Meta Superintelligence Labs, a frontier-model org led by Alexandr Wang (former Scale AI CEO, now Meta's Chief AI Officer) and Nat Friedman (ex-GitHub CEO), with Shengjia Zhao (ex-OpenAI, ChatGPT co-creator) named chief scientist. The 10-K frames the mission explicitly: "advance our vision to deliver personal superintelligence for everyone…AI that surpasses human intelligence".
• The value side (AI ad ranking + Advantage+): AI powers content ranking, the discovery engine (recommendations), and the ad delivery/targeting/measurement stack. Management states AI investments "support initiatives across our products…the tools advertisers use to reach customers". Meta's end-to-end AI ad solutions (Advantage+ suite) crossed a >$60B annual revenue run-rate as of Q3-2025, with 4M+ advertisers using its generative-AI ad tools.

Customers: Advertisers (the substantial-majority revenue source) — "as is common in the industry, our marketers do not have long-term advertising commitments with us". This is the central contract-structure fact: no take-or-pay, no recurring lock-in — revenue is re-won every quarter via auction. There is no single-customer concentration (millions of advertisers), which is a strength, but the demand is cyclical and macro-sensitive.

Suppliers: NVIDIA (GPUs), Broadcom (custom MTIA accelerators — Gen 2 now in production ), TSMC (silicon), the cloud/colocation providers behind $237.67B of non-cancelable contractual commitments, and energy providers (3-25 year clean-energy purchase agreements ).

Competitors: OpenAI, Google/DeepMind (Gemini), Anthropic (Claude), xAI (Grok) on the frontier-model axis; Google and Amazon on the ad-and-recommendation axis; ByteDance/TikTok on attention.

Lens 2 · Supply Chain

Upstream inputs → Meta → end customer, named at each link:

Silicon / accelerators Infrastructure Models & products Monetization → end customer
───────────────────── ────────────── ────────────────── ──────────────────────────
NVIDIA (H100/H200/GB200) → Self-built data centers → Llama (open), Muse Spark → Advantage+ ad suite → ~4M+ advertisers
TSMC (fab for MTIA + NVDA)→ Louisiana DC "Venture" → (closed frontier), FAIR → AI ranking / discovery → 3.56B DAP (the audience)
Broadcom (MTIA Gen 2 ASIC)→ ($27B est. cost, 20% Meta) research, Avocado/Mango → WhatsApp/Messenger (eyeballs sold to advertisers)
Energy (3-25yr clean PPAs)→ $237.67B contractual (open-weight, 2026) Business AIs (10M conv/wk)
 commitments (cloud+infra)

Chokepoints / single-source dependencies:

1. NVIDIA GPU allocation — the binding constraint on training-cluster scale; Meta diversifies with in-house MTIA silicon (Broadcom-partnered, Gen 2 in production ), but inference/training at frontier scale still leans on NVIDIA.
2. Power and land — the genuinely scarce input. The Louisiana data-center Venture (entered Oct 2025, 20% Meta interest, ~$27B total estimated development cost, ~$12.31B initial lease commitment, ~$28B residual-value-guarantee threshold, max exposure $45.99B) is an off-balance-sheet VIE structure built specifically to "meet future infrastructure capacity needs as AI markets…develop". This is Meta financing power/shells through an unconsolidated entity — a structural tell that the buildout has outgrown the balance sheet's comfort zone.
3. Cloud capacity — $14.72B of contingent cloud-capacity purchase obligations over five years (reducible if the provider resells), plus a further ~$24B of infrastructure commitments signed in April 2026 alone.

The chain is not generic — every link has a name. The constraint has migrated from chips (2023-24) to power + capital (2025-26).

Lens 3 · Competitive Advantages (moats)

The AI franchise inherits the parent's moats and adds one of its own.

• Distribution moat (the decisive one): 3.56B daily active people. Meta does not need to acquire AI users — it ships Meta AI into Facebook, Instagram, WhatsApp, Messenger to a captive 3.5B audience. OpenAI must buy distribution; Meta owns it. This is the asymmetry bulls underrate.
• Data + feedback moat: proprietary engagement and conversion data trains ranking/recommendation models no competitor can replicate. The ad-AI flywheel (better models → better targeting → higher ROAS → more ad budget → more data) is the moat that already monetizes.
• Capital moat: $115.80B operating cash flow in FY2025 lets Meta self-fund a $115-145B capex year without external financing pressure that would cripple a pure-play. Only ~5 companies on earth can fund this.
• Bargaining power: Over advertisers — high (millions of fragmented buyers, no single one matters; auction sets price). Over suppliers — mixed: strong vs. most, but weak vs. NVIDIA (the one supplier with pricing power over Meta), which is precisely why MTIA exists.

The moat that broke: the open-weight Llama strategy as a competitive moat. Llama 4 Behemoth (≈2T params) was previewed April 2025, then effectively shelved — never shipped as public weights, demoted to an internal "teacher model" after mid-training MoE-routing/chunked-attention problems. Meta then released Muse Spark (April 2026) — its first closed-weight frontier model, built by MSL — a strategic reversal from "open source is the moat." Open-source Avocado/Mango variants are promised for 2026, but the narrative that Meta would commoditize the model layer to its own advantage has visibly stalled. The moat is now distribution + data + capital, not model leadership.

Lens 4 · Segments

Meta reports two segments: Family of Apps (FoA) and Reality Labs (RL). The AI unit's costs are absorbed almost entirely within FoA (R&D + infrastructure); RL is the metaverse/hardware money-loser. All numbers ``.

By segment — revenue & operating income (FY, $M):

Read: FoA operating income grew +18% YoY (87.1→102.5B) — the engine is accelerating, not decelerating. RL loss widened to $19.19B (and management guides FY2026 RL losses "similar to 2025" and FY2026 Q1 RL loss was $4.03B ). RL is a ~$19B/yr tax on the franchise that has nothing to do with the AI thesis — but note neural interfaces/AI glasses increasingly are AI-adjacent.

By geography (FY2025 rev, $M): US & Canada 78,866 (US alone 74.78B); Europe 46,569; Asia-Pacific 53,817; Rest of World 21,714. US/Canada ~39% of revenue — the highest-ARPP, most-defensible base.

Q1-2026 acceleration: FoA revenue $55.91B (+33% YoY), FoA op income $26.90B (+24% YoY); RL revenue $402M, RL loss $(4.03)B. Ad impressions +19%, price-per-ad +12% YoY — both volume and price rising, which is the signature of AI-improved ad relevance pulling demand. There is no "GenAI revenue" segment line — the AI payoff shows up inside FoA advertising as higher impressions × higher price, not as a separate disclosed number.

---

Phase B — Measure performance

Lens 5 · Earnings Result (Q1-2026, the latest print)

The Q1-2026 print (reported Apr 29, 2026):

Beat/miss: Revenue ($56.31B) beat consensus; the headline +61% net-income jump overstates operating performance — it was flattered by an $8.03B discrete CAMT tax benefit (U.S. Treasury Notice 2026-7, partially reversing the Q3-2025 OBBBA charge). Operating income (+30%) and revenue (+33%) are the clean signals. The negative $(1.12)B interest-and-other line (vs. +$827M prior year) is the other swing factor — consistent with marks on non-marketable equity (Scale AI / the Venture) and rising interest expense from the Nov-2025 debt issuance.

What drove it: advertising — ad impressions +19%, price-per-ad +12%. Margin held in the high-30s/40% despite R&D +46% — the ad engine is absorbing the AI investment so far.

Balance-sheet flags: FY2025 close — cash + marketable securities $81.59B, long-term debt $58.74B (up from a near-debt-free posture after $29.91B notes issued Nov 2025). Q1-2026: $5.00B of money-market funds reclassified to restricted cash against a multi-year purchase agreement — a small but telling sign of capex commitments tying up liquidity.

Market reaction (the most important fact in this lens): despite the beat, the stock fell ~7-10% post-print — the worst price reaction in Meta's last six earnings reports — entirely on the capex guidance raise. The market is no longer rewarding the beat; it is punishing the spend. That is a regime change in how META trades.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts/ on the shelf — this lens is ``. Sentiment arc across the last ~4 cycles:

• 2023 ("Year of Efficiency"): cost discipline, 21k layoffs, margins 25%→42%, stock tripled. Tone: austerity rewarded.
• 2024-25: pivot from efficiency to aggressive AI investment; Zuckerberg reframes capex as a bet you'd rather over- than under-make.
• Q3-2025 → Q1-2026: tone shifts to conviction-under-fire. Management defends the $125-145B capex as supporting "AI efforts and core business" and leans on adoption proof points (Advantage+ >$60B run-rate, 4M+ GenAI advertisers, 10M weekly Business-AI conversations ).

Recurring phrases: "personal superintelligence," "core business," "operating efficiently while investing in significant opportunities". What they stopped saying: "Year of Efficiency" is gone; the open-source-leadership framing is muted (post-Behemoth, post-Muse-Spark). Tension signal: FT reporting of friction between Zuckerberg and Wang (Wang reportedly finds the micromanagement "suffocating") — a culture/retention risk inside the very lab that cost billions to assemble.

Lens 7 · Comps

Peer multiples are ``, dated; the _index.json holds no peers, so these are web-sourced megacap AI names. Spot/valuation figures conflict across sources (see note) — labeled where used.

Read: META is the cheapest of the megacap AI cohort on forward earnings (~17-19x vs. NVDA ~21x), despite +33% revenue growth. The market is applying an AI-spender discount, not an AI-winner premium — the inverse of how it treats NVIDIA. Dividend yield is small (Meta pays a token dividend, $5.32B total dividends FY2025 ); 5-yr average ROE is not cleanly sourced this pass (n/a), but FY2025 net income $60.46B on a large equity base implies ROE comfortably in the 25-35% zone. I am not fabricating GOOGL/MSFT multiples — they were not returned with a sourced date this pass.

Lens 8 · Stock-Price Catalysts (>5% moves, ~5yr pattern)

What actually moves META (`` + filing context):

• Feb 2022 — −26% in a day: the "Year of Efficiency" precursor crash — first-ever DAU decline + Apple ATT hit to ad targeting. Lesson: the market punishes user-growth stalls + ad-targeting threats hardest.
• 2023 — repeated +20%+ moves: "Year of Efficiency" earnings beats; margins 25%→42%; stock tripled. Lesson: cost discipline + margin expansion is the single most powerful up-catalyst.
• 2024-2025 — grind to high-$700s: AI-driven ad-targeting reaccelerated growth. Lesson: ad reacceleration > everything when paired with discipline.
• Q1-2026 — −7-10%: capex guide raised to $125-145B; worst reaction in six prints. Lesson: in 2026 the polarity flipped — capex raises now down-catalysts even on an EPS beat.

Pattern: META reacts to (1) user/engagement trend, (2) margin direction, and now (3) capex-vs-ROI perception — far more than to the absolute EPS number. The Q1-2026 reaction proves the market has re-weighted toward (3): it wants to see the AI return, not just the AI spend.

---

Phase C — Judge people & books

Lens 9 · Management

• Mark Zuckerberg (CEO/CODM, founder). Track record: built a ~$200B-revenue, ~$83B-operating-income machine; executed the 2023 efficiency turn that tripled the stock; called the AI-ranking bet early. Skin in the game: controls a majority of voting power via Class B super-voting shares (342.4M Class B vs. 2,196M Class A ) — the 10-K flags this as a risk factor: "Our CEO has control over key decision making". This is the central governance fact: the $135B/yr superintelligence bet is one founder's conviction with no board check.
• Capital-allocation history: Reinvest-heavy. FY2025 — $72.22B capex, $26.26B buybacks, $5.32B dividends, $18.33B purchases of non-marketable equity (Scale AI $13.80B closed in 2025; Jio $5.82B). The Scale AI stake is the financial expression of the talent raid (Wang came with it). Zuckerberg's record is bimodal: brilliant when reinvesting in the core (ad AI), value-destructive when reinvesting in the periphery (RL: ~$70B cumulative losses 2023-25 ).
• AI leadership: Alexandr Wang (Chief AI Officer), Nat Friedman, Shengjia Zhao. Assembled at extreme cost — Sam Altman publicly cited $100M signing bonuses; reports of a $1.5B package for a Thinking Machines Lab hire. Retention risk is live (Zuckerberg-Wang friction ).
• Red flags: founder-control with no governance brake; comp inflation distorting SBC; the Louisiana VIE structure (off-balance-sheet capex). Archetype: founder-operator with absolute control — high-variance, currently betting the franchise's free cash flow on a thesis (superintelligence) with no proven monetization path yet.

Lens 10 · Forensic Red Flags

Accounting-risk scan, every figure labeled.

• Useful-life change (the one to watch): In Jan 2025 Meta extended the estimated useful lives of "most servers and network assets to 5.5 years," cutting FY2025 depreciation by $2.92B and lifting net income by $2.59B (+$1.00 diluted EPS). This is a legitimate but earnings-flattering estimate change — and it runs directly counter to the AI-hardware reality, where GPU generations obsolete in 2-3 years. Stretching depreciation to 5.5 years while spending $135B/yr on rapidly-obsolescing accelerators is the single biggest forensic flag: it defers a depreciation wave that will eventually hit the income statement hard. Watch for a future reversal (shortening lives → a depreciation step-up).
• Tax volatility / one-offs: FY2025 net income ($60.46B) was below FY2024 ($62.36B) despite higher operating income ($83.28B vs $69.38B) — entirely because of the $15.93B OBBBA discrete charge (incl. $14.03B deferred-tax valuation allowance) in Q3-2025. Q1-2026 then booked an $8.03B CAMT benefit reversing part of it. GAAP net income is noisy right now — use operating income and FCF to judge the business.
• Cash flow vs. earnings: FY2025 OCF $115.80B vs. net income $60.46B — OCF exceeds earnings (healthy; depreciation + SBC add back). But FCF fell to $43.59B from $52.10B YoY purely because capex jumped to $69.69B (PP&E) — capex is now consuming the FCF growth. This is the number that converts to a bear case if revenue growth slows.
• Off-balance-sheet exposure: the Louisiana Venture — max exposure $45.99B, not consolidated (Meta deemed not the primary beneficiary). Plus $237.67B non-cancelable commitments and $182.88B of leases not-yet-commenced. The true capital footprint is far larger than the balance sheet shows.
• SBC: large (employee compensation is the biggest cost line in both segments) and rising with the AI talent war; the 10-K explicitly warns of "substantial additional share-based compensation expense and dilution" from AI hiring. Buybacks ($26.26B FY2025) partly offset dilution — diluted shares ~2,574M FY2025, roughly flat-to-down vs prior years.

Regulatory findings (required sub-section):

• SEC LR / AAER: None. "No LR found" and "No AAER found" for the company 2021-2026 per EDGAR EFTS.
• Non-SEC enforcement (``): Material and AI-specific. (1) FTC 6(b) inquiry (Sept 11, 2025) into AI "companion" chatbots — orders issued to Meta, Alphabet, OpenAI, xAI, Snap, Character.AI on minors' safety/COPPA compliance, triggered by an Aug-2025 Reuters investigation revealing an internal Meta standards doc that permitted bots to have "romantic or sensual" chats with children; Meta responded with new teen AI parental controls (Oct 2025). (2) EU DSA formal proceedings (opened May 16, 2024) on systemic risks to minors on Instagram/Facebook. (3) Congressional + state-AG inquiries into Meta AI chatbots.
• 10-K Item 3 / Legal Proceedings (``): Two material clusters bear directly on the AI thesis: (a) AI training copyright — Kadrey et al. v. Meta (consolidated): the court granted Meta summary judgment on fair use as to the named plaintiffs (June 25, 2025), but the distribution claim survives (alleged distribution of books during the torrent/download process), with SJ motions set for July 16, 2026; further cases (Entrepreneur Media, TED, etc.) filed Nov 2025, some headed to trial mid-2027. Statutory copyright damages are per-work — "may result in substantial damages, particularly given the large volumes of data required to train AI models". (b) Youth-safety / social-media-addiction MDL + state-AG cases — first personal-injury trial began Jan 27, 2026 (LA Superior Court); NM AG trial scheduled Feb 2, 2026; MDL school-district bellwether June 15, 2026; plaintiffs seek damages "up to the high tens of billions of dollars"; 100,000+ mass-arbitration demands re Instagram.
• IRS transfer pricing: Tax Court opinion (May 22, 2025) valued transferred IP at $7.79B ($1.48B above Meta's figure); a 2017-2019 Notice (Sept 2025) asserts an additional $15.89B in tax plus interest/penalties. Gross unrecognized tax benefits $17.82B (Q1-2026).

Net: clean on SEC accounting enforcement; the real legal/regulatory risk is AI-specific (copyright training liability + companion-chatbot/youth-safety) and tax (transfer pricing) — material but, in the firm's $200B-revenue context, manageable in the aggregate per management's own framing.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (FY2026 / FY2027 / FY2028 EPS)

Built bottom-up from FY2025 actuals + the latest print + guidance. Anchors: FY2025 diluted EPS $23.49; Q1-2026 diluted EPS $10.44 (incl. ~$3.20/sh from the one-off CAMT benefit ); FY2026 capex guide $125-145B, effective tax rate guide 13-16% for the rest of 2026; RL loss FY2026 "similar to" the FY2025 $19.19B; diluted share count ~2,560M, roughly flat (buybacks ≈ SBC dilution).

Input lines (FY2026 base):

• Revenue: Q1 ran +33%; assume full-year +22-25% as comps stiffen and macro normalizes → ~$245-250B.
• Operating margin: held ~40% in Q1 despite R&D +46%; assume ~38-40% full-year as depreciation from the capex wave builds → op income ~$94-98B.
• Tax: low — 13-16% guided range, plus the $8.03B Q1 discrete benefit already banked.
• Shares: ~2,560M diluted.

All EPS figures `` — arithmetic: FY2026 base ≈ ($248B rev × 38.5% op margin − $3B net interest) × (1 − 0.15) ÷ 2,560M sh ≈ **$30-32**. Consensus base case from the street clusters $30-34, consistent with my base. Note the FY2026 number is artificially aided by the ~$3.20/sh one-off tax benefit — normalized FY2026 EPS is closer to ~$29-30.

(Per --watchlist rules, NOT logging a forecast.ts Brier forecast — breadth loop.)

Lens 12 · Bull vs Bear

Bull case. Meta owns the one thing every AI lab is bleeding cash to rent: distribution to 3.5B people, plus the only AI use-case that already prints money at scale — ad ranking. The Advantage+ suite is past a $60B run-rate and ad impressions and price are both rising +double-digits — direct evidence the AI capex is already converting to revenue inside FoA. Operating income grew +18% (FY) and FoA +24% (Q1) — this is not a company "waiting" for AI ROI; it's harvesting it in the core while the moonshot (superintelligence) is a free call option funded by $116B of OCF. At ~17-19x forward earnings for +22-25% growth, you are paying a discount multiple for the best-capitalized AI franchise in the world. The contrarian view: the market has priced META as the AI loser precisely when the ad-AI flywheel is inflecting — a textbook expectations/reality gap.

Bear case. Three things could permanently impair the thesis: (1) the depreciation wave — $135B/yr of 2-3-year-half-life GPUs being depreciated over a stretched 5.5-year schedule means a multi-year depreciation step-up is deferred, not avoided; when it lands (and possibly resets shorter), margins compress structurally and FCF (already down to $43.6B from $52.1B ) gets squeezed hard. (2) Single-monetization-vehicle risk — the entire $135B infrastructure base is underwritten by one revenue engine (advertising); Business AIs hit 10M weekly conversations but are not yet monetized, so the incremental superintelligence spend has no incremental revenue line yet. (3) The open-weight thesis is dead — Behemoth shelved, Muse Spark closed; Meta is now a follower in frontier models spending frontier-leader money. Pre-mortem (18 months out, thesis broke): capex hit $145B+, ad growth decelerated to ~10% as macro softened, depreciation lives were shortened back toward 3-4 years, FCF halved, and the market re-rated to ~14-16x → stock back to the ~$540 zone. Are multiples too high? No — they're low; the risk is earnings, not the multiple.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration: ~98% of revenue is advertising, and advertisers have no contractual commitment — every dollar is re-auctioned quarterly. A 2008/2022-style ad recession + the $135B capex commitment = a vicious operating-leverage reversal. The buildout is fixed; the revenue is cyclical.
• The most dangerous competitor bulls underrate: not OpenAI — Google. Google has the same distribution-at-scale (Search, YouTube, Android, Chrome) and a credible frontier model (Gemini) and its own TPU silicon (no NVIDIA tax). If the AI-assistant layer cannibalizes the open web and Google wins the assistant, Meta's ranking edge erodes from both ends. Meta's frontier-model stumble (Behemoth) means it is behind on the one capability that future ad/recommendation moats may require.
• Worst capital-allocation moves: ~$70B cumulative RL losses 2023-25 with $402M/qtr revenue — a >40:1 burn-to-revenue ratio that the founder's voting control makes uncheckable. The $13.8B Scale AI stake + $100M-$1.5B individual comp packages are an unproven talent bet with live retention risk (Zuckerberg-Wang friction ).
• Accounting: the 5.5-year server-life extension flattered FY2025 EPS by $1.00 — aggressive given AI-hardware obsolescence; a reversal is a when, not an if.
• What must hold for $600: ad growth stays ≥20%, margins stay high-30s, the depreciation schedule isn't shortened, and no adverse copyright verdict in Kadrey's distribution claim (July 2026) or a "high-tens-of-billions" youth-safety judgment. If growth disappoints 20-30% (rev to +12-15%, EPS to ~$27 ), the stock is worth ~$430-490 at 16-18x — a 20-30% drawdown.
• Single scenario that permanently impairs: an AI-assistant paradigm shift (ChatGPT/Gemini becomes the default interface) that structurally reduces time-spent in Meta's feeds → fewer impressions → the ad engine that funds everything contracts while $135B/yr of committed capex keeps running. Plausibility: moderate — the 3.5B DAP moat is sticky, but younger-user engagement erosion is a disclosed risk.

Lens 14 · Management Questions (ordered by information value)

1. Of the $125-145B FY2026 capex, how much is training superintelligence vs. inference/serving the existing ad-AI and recommendation workloads — and what is the incremental revenue per dollar of each bucket?
2. You extended server useful lives to 5.5 years in Jan 2025. Given GPU generational cadence, under what conditions would you shorten them again, and what would the depreciation step-up be?
3. Business AIs hit 10M weekly conversations — what is the concrete monetization model and timeline, and what ARPP do you underwrite for it?
4. Llama 4 Behemoth was shelved and Muse Spark is closed-weight. Is open-weight still a strategic pillar, or has the strategy permanently shifted to closed frontier + selective open releases?
5. What return threshold (ROIC, incremental ad revenue) would cause you to cut AI capex, and what would you need to see to raise it again beyond $145B?
6. The Louisiana Venture carries $45.99B max exposure off-balance-sheet. Why the unconsolidated VIE structure rather than direct ownership, and how many more such structures are planned?
7. How do you retain the MSL leadership (Wang, Friedman, Zhao) given the reported internal friction and the comp packages that recruited them — what's the vesting/retention design?
8. If U.S. ad growth decelerates to ~10% in a downturn, which capex commitments are cancelable, and what is the true fixed vs. flexible split of the $237.67B in commitments?
9. What is your contingency if the Kadrey distribution claim (July 2026 SJ) or the AI-training copyright cases produce a per-work statutory-damages verdict at scale?
10. How do you quantify the ad-revenue uplift specifically attributable to AI ranking improvements vs. price/impression growth that would have happened anyway?
11. With RL still losing ~$19B/yr, at what point does the AI-glasses/neural-interface roadmap justify the spend, and how is that decision insulated from sunk-cost bias?
12. What share of the 3.5B DAP base now interacts with Meta AI weekly, and is that engagement additive to time-spent or substituting for feed time that monetizes better?
13. How exposed is the training/inference roadmap to a single NVIDIA supply constraint, and what % of 2026-27 compute will run on MTIA?
14. Given your majority voting control, what governance mechanism (if any) checks a capex decision the board disagrees with?
15. What is the youth-safety litigation reserve, and how do you bound the "high tens of billions" exposure cited by plaintiffs across the MDL + state-AG cases?

---