← back to ranking

PINS · Pinterest, Inc.

Internet Content & Information · mkt cap $13.7B · calls: Q1 FY2026 vs Q4 FY2025
62.0 conviction · conf-adj 62

conf 4/10 partial

enthusiasm:27.0 · trend:8 · quantifies:0 · impact:0 · under_radar:14 · credibility:5 · business_impact:8 · disruption:-6 · commitment:6 · confirmation:0

Enthusiasm latest 9 / prev 8 (rising)

Pinterest frames AI as proprietary taste-graph-driven personalization (PinRack, search ranking, Canvas) plus an automated ads stack (Performance Plus, ROAS/bidding models, measurement-linked optimization) that is already tied to engagement and advertiser KPIs, with unusually dense quantification for a consumer internet name. Enthusiasm is rising: Q1 adds site-wide PinRack rollout, Canvas for ad creative, and explicit revenue attribution of AI bidding to large-retailer recovery, while Q4 established the model stack (OmniSage, PinFM, Navigator One, Assistant beta). Credibility is moderate-to-strong on product metrics and pilot/adoption data, but weak on isolating AI's direct revenue contribution—management still attributes monetization gaps primarily to go-to-market and measurement execution rather than user-side AI performance.

GROUNDED NEXT-FY IMPACT vs CONSENSUS

Grounded on actual base — revenue $4.2B · net income $0.4B · net margin 9.9% · diluted EPS 0.61

These are next-fiscal-year annual uplift estimates, not next-quarter numbers.

Aggregate next-FY est. rev uplift: % · next-FY EPS uplift: % · vs analysts: inline · priced in: low (model's call-read: high; verdict above is the hard-data one used for ranking) · confidence: 4/10

ClaimFigureArithmeticNext-FY Rev %Next-FY EPS %
PinRack search fulfillment +180bp
engagement · soft
~180 bps180 bps = 1.8% lift in search fulfillment (product/engagement metric), not revenue. No disclosed mapping from fulfillment to monetized ad/search revenue, so cannot convert to a $ uplift without inventing elasticity.
PinRack CPA/CPC -180bp for advertisers
engagement · soft
~180 bps reduction-1.8% price per click/action improves advertiser efficiency (could lift spend) but volume offset undisclosed and no disclosed share of revenue on CPC/CPA pricing; unsizable.
Search ranking update fulfillment +70bp
engagement · soft
~70 bpsEngagement metric (1.7% fulfillment); no revenue base disclosed.
Search ranking update saves +390bp
engagement · soft
~390 bpsSaves is an engagement metric (3.9%); not mapped to a revenue line in the inputs.
User context window 30x in search ranking
other · soft
30-foldTechnical capability metric, no financial anchor.
16k user actions over 2yr for search ranking
other · soft
up to 16k actionsInput-scale descriptor only; no financial translation.
Canvas order-of-magnitude lower cost
cost · soft
~10x lower cost vs 3rd-party (~90% savings)Real inference-cost saving but no disclosed $ third-party/GPU spend base, so cannot compute after_tax_saving or eps_uplift_pct.
Performance Plus = ~30% of lower-funnel revenue
engagement · soft
~30% shareA revenue-MIX share of lower-funnel revenue, not an incremental uplift; lower-funnel revenue base not disclosed in $, so no incremental computable.
PP adopters grew lower-funnel spend ~2x non-adopters
engagement · soft
~2x rateRelative growth rate only; baseline growth and adopter LF spend base undisclosed; unsizable.
Mejuri PP A/B: +46% ROAS
engagement · soft
+46% ROASSingle-advertiser four-week test; no spend weight in PINS revenue; not generalizable.
Mejuri PP A/B: +62% conversions
engagement · soft
+62% conversionsSingle-advertiser anecdote, unweighted; unsizable to revenue.
Shopping ROAS model gains up to 11%
engagement · soft
up to +11% ROASExperimentation subset, advertiser-side efficiency; no disclosed shopping/LF revenue $ or rollout phasing to anchor.
LTV ROAS +15-20% (measurement-linked bidding pilot)
engagement · soft
+15-20% LTV ROASN=1 early pilot; no budget share disclosed; unsizable.
CTV incremental reach +190% (TV Scientific)
engagement · soft
+~190% reachOne early partner outcome metric, not PINS recognized revenue; no PINS booking/rev share disclosed.
CTV incremental sales +159% (TV Scientific)
engagement · soft
+159% salesEarly partner anecdote (advertiser's sales, not PINS revenue); not sized to FY2026.
PP requires half as many setup inputs
other · soft
half the inputsUX/adoption-friction metric; no $ opex or revenue conversion disclosed.
TV Scientific outcomes +27% (taste graph)
engagement · soft
+27% outcomesOff-platform partner outcome metric; PINS revenue attribution/consolidation timing not quantified for FY2026.
TV Scientific purchases +65% (taste graph)
engagement · soft
+65% purchasesEarly partner purchase metric, no PINS revenue bridge; unsizable.
GPU/cost-of-revenue 'modest' headwind FY26 (qualitative)
cost · soft
'modest headwinds'Qualitative version of the quantified ~100bp item; not separately sized to avoid double count.
OmniSage site-wide saves +450bp
engagement · soft
+450 bps saves4.5% saves (site-wide engagement) metric, no disclosed saves-to-revenue elasticity or base.
PinFM site-wide saves +240bp
engagement · soft
+240 bps saves2.4% engagement metric, no revenue base disclosed.
Navigator One ~90% cost reduction vs 3rd-party
cost · soft
~90% cost cutReal inference-cost saving but no disclosed dollar/volume base; cannot size $.
Pinterest Assistant: +25pp commercial-question share
engagement · soft
+25pp shareEngagement/mix metric on assistant queries; no disclosed revenue base to convert query-mix shift into monetized dollars.

Assumptions: Tax rate 21%. EPS uplift sized against the NON-GAAP/adjusted earnings basis consensus uses: FY2025 consensus net income $1,190.4M (EPS $1.649) — NOT GAAP NI $416.9M / EPS $0.61, which is depressed by SBC/other items and would yield a meaningless inflated %. No incremental-margin assumption was needed because no topline claim carried a dollar-anchored base to flow to revenue. Cost headwind sized on current revenue ($4,221.8M) per method; on FY2026 consensus revenue ($4,861.8M) it would be ~$48.6M / -3.2% EPS — same order. Phasing: the ~100bp cost headwind is explicit FY2026 guidance. Side: 100% adopter — every claim is Pinterest improving its own ad products, engagement, search, or infra costs; PINS sells no AI compute/chips/capacity to third parties.

Top line: All AI topline evidence (search fulfillment +70-180bp, saves +240-450bp, Performance Plus = ~30% of lower-funnel revenue and adopters growing spend ~2x non-adopters, ROAS lifts of 11-46%, CTV +159-190%) is genuinely the engine behind the business — but management quantified it only as engagement/relevancy basis points and per-advertiser anecdotes, never anchoring a single one to a revenue-dollar base or segment size. None are independently sizable into a revenue %, so aggregate adopter rev_uplift = null. The mechanism is real; the dollar magnitude is undisclosed.

Bottom line: The ONLY dollar-anchorable AI item is the GPU/cost-of-revenue investment: ~100bp of revenue = ~$42M, ~$33M after-tax, i.e. roughly -2.8% against the adjusted ~$1,190M earnings base — a near-term EPS HEADWIND. The offsetting cost wins (Canvas ~10x cheaper, Navigator One ~90% cheaper, ~50% AI-generated code, half of OpEx savings reinvested) are all real but undisclosed in dollars, so the only quantifiable bottom-line effect of the disclosed AI program in FY26 is a modest margin drag from the buildout, not a margin tailwind.

[impact n/m (all claims soft/unanchored)] Consensus already embeds the AI flywheel: FY26 revenue $4,861.8M vs FY25 $4,221.8M = +15.2%, and adjusted EPS $1.899 vs $1.649 = +15.2%, with EPS growth holding flat to revenue despite the explicitly-guided ~100bp gross-margin headwind. That headwind is in management's own guidance, so the -2.8% cost drag is already in the numbers. Because no AI topline claim was dollar-anchored, there is no quantified figure that points clearly above the +15% consensus trajectory — the gap is zero on the math we can actually do. The bull case rests on undisclosed magnitudes (Performance Plus multiyear ramp, TV Scientific/full-funnel CTV), which are upside optionality, not a measurable consensus miss.

MODEL CONSENSUS (impact)

partial

Both independently mark every claim soft=true with null rev/eps pcts (no disclosed $ bases); only claim-type labels and granularity differed.

Conflicts reconciled
FieldOpus 4.8GPT-5.5
Rev uplift %
EPS uplift %-2.8
Priced inhigh
vs analystsinline
Confidence4
Top lineAll AI topline evidence (search fulfillment +70-180bp, saves +240-450bp, Performance Plus = ~30% of lower-funnel revenue and adopters growing spend ~2x non-adopters, ROAS lifts of 11-46%, CTV +159-190%) is genuinely the engine behind the business — but management quantified it only as engagement/relevancy basis points and per-advertiser anecdotes, never anchoring a single one to a revenue-dollar base or segment size. None are independently sizable into a revenue %, so aggregate adopter rev_uplift = null. The mechanism is real; the dollar magnitude is undisclosed.
Bottom lineThe ONLY dollar-anchorable AI item is the GPU/cost-of-revenue investment: ~100bp of revenue = ~$42M, ~$33M after-tax, i.e. roughly -2.8% against the adjusted ~$1,190M earnings base — a near-term EPS HEADWIND. The offsetting cost wins (Canvas ~10x cheaper, Navigator One ~90% cheaper, ~50% AI-generated code, half of OpEx savings reinvested) are all real but undisclosed in dollars, so the only quantifiable bottom-line effect of the disclosed AI program in FY26 is a modest margin drag from the buildout, not a margin tailwind.
ReasoningConsensus already embeds the AI flywheel: FY26 revenue $4,861.8M vs FY25 $4,221.8M = +15.2%, and adjusted EPS $1.899 vs $1.649 = +15.2%, with EPS growth holding flat to revenue despite the explicitly-guided ~100bp gross-margin headwind. That headwind is in management's own guidance, so the -2.8% cost drag is already in the numbers. Because no AI topline claim was dollar-anchored, there is no quantified figure that points clearly above the +15% consensus trajectory — the gap is zero on the math we can actually do. The bull case rests on undisclosed magnitudes (Performance Plus multiyear ramp, TV Scientific/full-funnel CTV), which are upside optionality, not a measurable consensus miss.

Rows highlighted where the two models disagreed.

QUANTIFICATIONS
Search fulfillment improvement (PinRack site-wide launch): approximately 180 basis points (Q1 FY2026, both)
“This launch improved search fulfillment by approximately 180 basis points.”
Advertiser CPA/CPC reduction (PinRack): roughly 180 basis point reduction (Q1 FY2026, topline)
“It also drove a roughly 180 basis point reduction in CPA and CPC for advertisers.”
Search fulfillment improvement (search ranking model update): approximately 70 basis points (Q1 FY2026, both)
“This launch improved search fulfillment by approximately 70 basis points and saves by approximately 390 basis points.”
Saves improvement (search ranking model update): approximately 390 basis points (Q1 FY2026, both)
“This launch improved search fulfillment by approximately 70 basis points and saves by approximately 390 basis points.”
User context window expansion in search ranking: 30-fold (Q1 FY2026, both)
“In Q1, we updated our proprietary search ranking model, extending user context windows within search by 30-fold, similar to the expansion we previously made to our home feed ranking model.”
User actions used for search ranking: up to 16 thousand user actions over a two-year period (Q1 FY2026, both)
“We now use up to 16 thousand user actions over a two-year period to inform the search results shown to each user.”
Canvas cost vs third-party models: an order of magnitude lower cost (current, bottomline)
“Canvas allows us to build experiences that reflect the high bar for visual quality and aesthetics that users and advertisers expect from Pinterest, Inc., while operating at an order of magnitude lower cost than leading third-party models.”
Performance Plus share of lower-funnel revenue: approximately 30% (Q1 FY2026 (~1 year after GA), topline)
“Just over a year in, approximately 30% of lower-funnel revenue is now running through Pinterest, Inc. Performance Plus campaigns, but we are still early in capturing the full opportunity, as adoption continues to expand and we continue to build out functionality of the suite.”
Performance Plus adopter lower-funnel spend growth vs non-adopters: nearly twice the rate (Q1 FY2026, topline)
“And importantly, in Q1, adopters of Pinterest, Inc. Performance Plus campaigns grew their lower-funnel spend nearly twice the rate of non-adopters.”
Mejuri ROAS lift (Performance Plus A/B test): 46% increase in ROAS (four-week A/B test, Q1 FY2026, topline)
“The Pinterest, Inc. Performance Plus campaign delivered a 46% increase in ROAS and a 62% increase in conversions, which led Mejuri to adopt Pinterest, Inc. Performance Plus campaigns more broadly.”
Mejuri conversions lift (Performance Plus A/B test): 62% increase in conversions (four-week A/B test, Q1 FY2026, topline)
“The Pinterest, Inc. Performance Plus campaign delivered a 46% increase in ROAS and a 62% increase in conversions, which led Mejuri to adopt Pinterest, Inc. Performance Plus campaigns more broadly.”
Shopping ROAS model experimentation gains: ROAS gains of up to 11% (Q1 FY2026 experimentation, topline)
“In experimentation, these improvements drove ROAS gains of up to 11% and are an indication of what continued investment in our ads platform can unlock.”
Lifetime value ROAS improvement (measurement-linked bidding pilot): 15% to 20% improvement in lifetime value ROAS (early testing, Q1 FY2026, topline)
“In early testing with one advertiser that prioritizes lifetime value, the advertiser cited a 15% to 20% improvement in lifetime value ROAS.”
CTV incremental audience reach (Pinterest audience data + TV Scientific): nearly 190% increase (early partner results, Q1 FY2026, topline)
“One early partner, a leading home furnishings omnichannel retailer, saw a nearly 190% increase in incremental audience reach and a 159% increase in incremental sales after leveraging Pinterest, Inc. audience data in its CTV campaigns.”
CTV incremental sales (Pinterest audience data + TV Scientific): 159% increase (early partner results, Q1 FY2026, topline)
“One early partner, a leading home furnishings omnichannel retailer, saw a nearly 190% increase in incremental audience reach and a 159% increase in incremental sales after leveraging Pinterest, Inc. audience data in its CTV campaigns.”
Performance Plus campaign setup inputs vs standard campaign: half as many inputs (current, both)
“In particular, we are focused on driving adoption of Pinterest, Inc. Performance campaigns, our automated bundle of bidding, budgeting, targeting, and creative features that reduces CPAs and CPCs while requiring half as many inputs to set up as a standard campaign.”
TV Scientific outcomes lift (Ready Q&A, taste graph on algorithms): 27% increase in outcomes (current/early, topline)
“That is one tangible example we talked about on the call of how we can use our data on top of algorithms to get even better outcomes and is part of what we are doing with AI models generally—both what we build in-house and where we retrain open-source models.”
TV Scientific purchases lift (Ready Q&A): 65% increase in purchases (current/early, topline)
“I would point you to what we are doing with TV Scientific. It is a very tangible example of what we can do with that data beyond our Pinterest, Inc. app, where we have been able to achieve a 27% increase in outcomes and a 65% increase in purchases by leveraging our taste graph on top of TV Scientific’s algorithms.”
GPU/cost-of-revenue investment headwind: modest headwinds from cost of revenue as a percentage of revenue in 2026 (FY2026, bottomline)
“Starting with cost of revenue, as with Q2, we continue to expect modest headwinds from cost of revenue as a percentage of revenue in 2026 as a result of the investments in areas such as additional GPU capacity, as well as the impact from the inclusion of TV Scientific.”
OmniSage site-wide saves lift: 450 basis point lift (2025 application, both)
“The application of OmniSage drove a 450 basis point lift in site-wide saves.”
PinFM site-wide saves lift: 240 basis point increase (2025 launch, both)
“This launch brought meaningful site-wide engagement gains, including a 240 basis point increase in saves across the platform.”
Navigator One cost reduction vs third-party model: approximately 90% reduction in cost (2025 framework, bottomline)
“This framework reduces latency and delivers approximately 90% reduction in cost versus utilizing a leading third-party proprietary model.”
Commercial question share lift with Pinterest Assistant vs text search: about 25 percentage points more (beta usage, Q4 FY2025 launch period, both)
“Compared with traditional text-based search, users are asking a significantly higher share of commercially oriented questions, about 25 percentage points more when using Pinterest Assistant.”
Internal code generation: roughly 50% of our new code is AI generated (current, bottomline)
“Lastly, AI is at the core of how we are improving efficiencies internally as roughly 50% of our new code is AI generated.”
Advertiser bid increase under measurement-linked optimization pilot: more than 30% (pilot to date, Q4 FY2025, topline)
“So far, this pilot has delivered promising results with one advertiser increasing its bids on Pinterest by more than 30%, reflecting the higher value it was seeing from the platform under this new value-based optimization approach.”
New customer conversions lift (Performance+ new customer acquisition beta): average of 64% (initial testing, topline)
“In initial testing, advertisers saw new customer conversions increase by an average of 64% in campaigns where new customer acquisition was enabled compared to control campaigns without it.”
Managed SMB monthly revenue growth rate vs non-adopters (Performance Plus): 12% higher monthly revenue growth rate (Q4 FY2025 / prior quarter reference, topline)
“So as we noted last quarter, we see a 12% higher monthly revenue growth rate with these managed SMB advertisers versus nonadopters.”
Multimodal visual search relevancy vs off-the-shelf models: 34 percentage points (latest models (prior calls reference), both)
“I shared on prior calls, that our latest multimodal visual search models outperform leading proprietary off-the-shelf models by 34 percentage points on the relevancy of shopping recommendations.”
Taste graph growth: nearly 40% (2025, both)
“In 2025, our taste graph grew by nearly 40% as users make more associations across pins, products, boards, retailers and brands.”
GPU capacity cost-of-revenue investment: approximately 100 basis points in 2026 (FY2026, bottomline)
“So we expect this cost of revenue investment to be approximately 100 basis points in 2026, similar to the gross margin outlook implied in my Q1 commentary earlier.”
OpEx savings reinvestment into AI talent: roughly half of those OpEx savings (FY2026, bottomline)
“Now we expect to reinvest roughly half of those OpEx savings primarily in our sales transformation and in AI talent.”
PAST (realized)
CURRENT (now)
FORWARD (guidance)
TRACK RECORD — PROMISE vs DELIVERY

67/100 track record   mixed  6 calls reviewed

Pinterest rarely sets numeric AI KPIs with deadlines; it mostly reports achieved model lifts. On discrete product milestones it usually ships (ROAS, Instacart, boards), but Assistant rollout and measurement expansion have slipped versus early timing language.

ROAS bidding live for all eligible advertisers by end of Q1 2025 — promised Q4 FY2024
delivered Q1 FY2025 reported ROAS bidding in general availability in March 2025.
Performance Plus creative automation (cropping, brightness, logo overlays) in H1 2025 — promised Q4 FY2024
partial Q1–Q2 FY2025 shipped resizing and beta creative-preview/generative tools, but not the full H1 feature set as framed.
Pinterest ads shoppable via Instacart soon — promised Q2 FY2025
delivered Q3 FY2025 launched Where to Buy and expanded Instacart-linked shoppable CPG flows in September.
Boards Made for You for UCAN users in coming weeks — promised Q3 FY2025
delivered Q4 FY2025 holiday go-to-market included personalized board/gift-guide experiences aligned with this launch.
Meaningfully broaden Pinterest Assistant access to U.S. users over coming months — promised Q4 FY2025
partial Q1 FY2026 still described Assistant as iterating in beta with no broad U.S. rollout yet.
Expand in-house measurement pilot to more large advertisers in H1 2026 — promised Q4 FY2025
partial Q1 FY2026 kept the pilot running but pushed broader expansion to later in 2026, not first-half timing.
PRICED-IN (REFINED)
LOW (room left)

Est. revisions falling  ·  Fwd P/E 12.5  ·  EV/Sales 3.3x

AI claim maps to United States And Canada, UNITED STATES, Europe

Analyst sentiment has cooled since early 2026 (combined strongBuy+buy fell from ~30 to ~19 while holds rose toward ~20) and consensus price targets stepped down from a ~$30.56 last-year average to ~$26.7, with only a marginal uptick month vs quarter. Forward valuation is not stretched at ~12.5x next-FY EPS and ~3.3x EV/Sales, so the market is not paying a premium for AI despite mid-teens revenue/EPS growth already in the forward curve. With no product segmentation, AI-driven ad monetization and efficiency would most plausibly show up in United States And Canada, UNITED STATES, and Europe revenue lines.
COVERAGE — ENTHUSIASM TRAJECTORY + CATALYSTS
8Q4 FY20248Q1 FY20257Q2 FY20258Q3 FY20259Q4 FY20259Q1 FY2026

AI enthusiasm across 6 calls — trend ↗ rising

From taste-graph ML metrics to named models, Assistant/Canvas launches, and quantified ad ROI plus Performance Plus adoption.

BUSINESS IMPACT - QUALITATIVE MATERIALITY

8/10 qualitative impact   material  near-term · mixed evidence

Where AI matters: visual discovery personalization and AI performance advertising

PinRack site-wide rollout, search-ranking lifts (+70-180bp fulfillment, +390bp saves), and Performance Plus (~30% of lower-funnel revenue with adopters spending ~2x non-adopters) show AI is actively driving engagement and ad efficiency, but management never dollar-anchors incremental revenue so upside magnitude stays inferred not proven.

Caveats: No isolated AI revenue or EPS uplift disclosed despite dense engagement metrics; ~100bp gross-margin headwind from AI/GPU infrastructure with undisclosed offsetting cost savings; Assistant broad rollout and measurement partner integrations have slipped versus prior timing; Chatbot-native discovery could compress session time if users shift to conversational shopping agents

AI DISRUPTION / CANNIBALIZATION RISK  two-sided · 4/10

Multimodal chatbots and AI shopping agents could substitute for Pinterest's visual browse-and-save discovery loop and cheap generative content could dilute curated aesthetic value, but proprietary taste-graph/save signals and AI-powered ad optimization have held user growth and monetization so far.

OPTIONS / MARKET STRUCTURE

option liquidity: good

ATM IV
TYPICAL BID-ASK
OPEN INTEREST

proxy inputs — dollar-ADV $395M · beta 0.923 · px $20.65

source: proxy (no options chain on FMP)
FMP /stable/ exposes no options-chain endpoint on this key, so ATM IV, bid-ask spread and open interest are unavailable. Liquidity below is a PROXY from dollar-ADV, beta and price level (a stand-in for option depth), not measured option-market data.

CONFIRMATION — INSIDERS · 13F · LANGUAGE
Mixed — insiders selling, institutions flat, management language 8/10 committed.
INSIDERS selling 10 open-market sell(s) vs 0 buy(s) — net distribution
INSTITUTIONS (13F) flat as of 2026-03-31: 116 new / 231 closed positions; 349 increased / 254 reduced; institutional ownership -5.23pp; -119 net 13F holders
MGMT LANGUAGE 8/10 committed AI framed with shipped metrics and causality; future rollout still qualified as early and over time.
commit “Ten straight quarters of double-digit user growth are the direct result of multiyear investments in AI improving personalization and curation within visual search and discovery.”
commit “This launch improved search fulfillment by approximately 180 basis points.”
commit “Just over a year in, approximately 30% of lower-funnel revenue is now running through Pinterest, Inc. Performance Plus campaigns”
VERBATIM AI QUOTES
“That is where our AI and proprietary taste graph come in.”
— William J. Ready, Q1 FY2026
“By understanding not just what a user is searching for today, but who they are and how their interests are evolving, we have made Pinterest, Inc. a highly personalized AI-powered shopping assistant.”
— William J. Ready, Q1 FY2026
“Second, keep AI at the core of everything we do, from powering our user experiences and ad platform to optimizing our internal operations.”
— William J. Ready, Q1 FY2026
“Ten straight quarters of double-digit user growth are the direct result of multiyear investments in AI improving personalization and curation within visual search and discovery.”
— William J. Ready, Q1 FY2026
“Every search, click, and save gives our AI more signal about who a user is and what they care about, which allows us to deliver more relevant and personalized experiences across the platform.”
— William J. Ready, Q1 FY2026
“In some cases, that means fit-for-purpose proprietary models that outperform leading third-party alternatives.”
— William J. Ready, Q1 FY2026
“In others, it means post-training suitable open-source models in our own environment within our cloud infrastructure that deliver comparable outcomes to third-party models, but at a fraction of the cost.”
— William J. Ready, Q1 FY2026
“An example of this is PinRack, our proprietary generative retrieval system, which is trained on user activity and our taste graph.”
— William J. Ready, Q1 FY2026
“Rather than building separate models optimized for each surface, PinRack is now a single model that generates personalized results for each user across all surfaces simultaneously, informed by the full depth of what we know about their taste and interests.”
— William J. Ready, Q1 FY2026
“We initially launched this model on search and related surfaces in 2025, and subsequently extended it in Q1 to serve content globally site-wide.”
— William J. Ready, Q1 FY2026
“This launch improved search fulfillment by approximately 180 basis points.”
— William J. Ready, Q1 FY2026
“It also drove a roughly 180 basis point reduction in CPA and CPC for advertisers.”
— William J. Ready, Q1 FY2026
“In Q1, we updated our proprietary search ranking model, extending user context windows within search by 30-fold, similar to the expansion we previously made to our home feed ranking model.”
— William J. Ready, Q1 FY2026
“We now use up to 16 thousand user actions over a two-year period to inform the search results shown to each user.”
— William J. Ready, Q1 FY2026
“This launch improved search fulfillment by approximately 70 basis points and saves by approximately 390 basis points.”
— William J. Ready, Q1 FY2026
“Our AI capabilities also extend into creative generation with Canvas, our in-house AI image generation model trained exclusively on Pinterest, Inc. data.”
— William J. Ready, Q1 FY2026
“Canvas allows us to build experiences that reflect the high bar for visual quality and aesthetics that users and advertisers expect from Pinterest, Inc., while operating at an order of magnitude lower cost than leading third-party models.”
— William J. Ready, Q1 FY2026
“It already supports Pinterest, Inc. Performance Plus creative optimization, enabling advertisers to dynamically edit backgrounds and transform basic catalog images into high-performing lifestyle images.”
— William J. Ready, Q1 FY2026
“With the newest version of the model now supporting real-time, high-fidelity image editing, particularly in key verticals, we expect to expand Canvas to enable more creative experiences for users and advertisers in the months ahead.”
— William J. Ready, Q1 FY2026
“Our AI investments are also translating into better advertiser performance, as Pinterest, Inc. Performance Plus, our AI-powered performance ad suite, continues to drive strong results for advertisers.”
— William J. Ready, Q1 FY2026
“In particular, we are focused on driving adoption of Pinterest, Inc. Performance campaigns, our automated bundle of bidding, budgeting, targeting, and creative features that reduces CPAs and CPCs while requiring half as many inputs to set up as a standard campaign.”
— William J. Ready, Q1 FY2026
“Just over a year in, approximately 30% of lower-funnel revenue is now running through Pinterest, Inc. Performance Plus campaigns, but we are still early in capturing the full opportunity, as adoption continues to expand and we continue to build out functionality of the suite.”
— William J. Ready, Q1 FY2026
“And importantly, in Q1, adopters of Pinterest, Inc. Performance Plus campaigns grew their lower-funnel spend nearly twice the rate of non-adopters.”
— William J. Ready, Q1 FY2026
“In Q1, we launched a native A/B testing tool in beta directly in Ads Manager, allowing advertisers to run structured, KPI-driven tests comparing Pinterest, Inc. Performance Plus campaigns to their existing ones.”
— William J. Ready, Q1 FY2026
“The Pinterest, Inc. Performance Plus campaign delivered a 46% increase in ROAS and a 62% increase in conversions, which led Mejuri to adopt Pinterest, Inc. Performance Plus campaigns more broadly.”
— William J. Ready, Q1 FY2026
“In Q1, we unified and retrained our Shopping ROAS models to better predict and optimize for advertiser return on ad spend across multiple stages of our ad stack.”
— William J. Ready, Q1 FY2026
“In experimentation, these improvements drove ROAS gains of up to 11% and are an indication of what continued investment in our ads platform can unlock.”
— William J. Ready, Q1 FY2026
“For our largest and most sophisticated advertisers, we are continuing to pilot integrations with their proprietary in-house measurement systems, which enables our bidding systems to respond dynamically to their specific definition of a successful outcome, whether that is customer lifetime value, profit per order, or something else entirely.”
— William J. Ready, Q1 FY2026
“In early testing with one advertiser that prioritizes lifetime value, the advertiser cited a 15% to 20% improvement in lifetime value ROAS.”
— William J. Ready, Q1 FY2026
“Whether an advertiser uses a first-party measurement system or a third-party partner, our goal is the same: help them better understand the full value Pinterest, Inc. is driving, while also helping us optimize our AI bidding systems toward the outcomes that matter most to them.”
— William J. Ready, Q1 FY2026
“We have already begun integrating Pinterest, Inc. audiences and signals with TV Scientific’s algorithms via TV Scientific’s buying platform.”
— William J. Ready, Q1 FY2026
“One early partner, a leading home furnishings omnichannel retailer, saw a nearly 190% increase in incremental audience reach and a 159% increase in incremental sales after leveraging Pinterest, Inc. audience data in its CTV campaigns.”
— William J. Ready, Q1 FY2026
“Over time, we expect to integrate TV Scientific capabilities directly into Pinterest, Inc. Performance Plus, turning Pinterest, Inc. into a full-funnel search, social, and CTV performance solution that should open larger and incremental budget pools.”
— William J. Ready, Q1 FY2026
“And we are evolving our sales incentive structures to drive more accountability and give a sharper insight into execution across the organization.”
— William J. Ready, Q1 FY2026
“We are also incorporating internal AI adoption and advertiser conversion signal quality into how we measure performance.”
— William J. Ready, Q1 FY2026
“We were the first major online platform to make accounts for users under 16 private only.”
— William J. Ready, Q1 FY2026
“We have also supported efforts like phone-free schools and App Store age verification while applying AI in ways to prioritize positivity.”
— William J. Ready, Q1 FY2026
“In closing, as AI reshapes how people discover, plan, and shop, Pinterest, Inc. is in a differentiated position.”
— William J. Ready, Q1 FY2026
“While these changes will take time to fully play out, we believe the progress we are making across the business and the outcomes from our AI investments will lead to durable growth over time.”
— Julia Brau Donnelly, Q1 FY2026
“The increase was primarily driven by Sales and Marketing due to headcount investments and marketing expenses, as well as R&D to support our AI and product initiatives.”
— Julia Brau Donnelly, Q1 FY2026
“However, AI-driven ad platform improvements, including bidding optimizations for this group, partially offset some of this headwind later in the quarter.”
— Julia Brau Donnelly, Q1 FY2026
“We anticipate Q2 2026 non-GAAP cost of revenue to grow sequentially from Q1 2026 by mid-single digits percent, partially driven by the full quarter impact from TV Scientific and our investment in GPU capacity.”
— Julia Brau Donnelly, Q1 FY2026
“Within R&D, we are continuing to invest in headcount to support our AI and product initiatives.”
— Julia Brau Donnelly, Q1 FY2026
“Starting with cost of revenue, as with Q2, we continue to expect modest headwinds from cost of revenue as a percentage of revenue in 2026 as a result of the investments in areas such as additional GPU capacity, as well as the impact from the inclusion of TV Scientific.”
— Julia Brau Donnelly, Q1 FY2026
“Importantly, we are already starting to see strong yield from our GPU capacity investments, including the engagement and performance improvements that William mentioned earlier.”
— Julia Brau Donnelly, Q1 FY2026
“Our user base is growing, our AI investments are producing measurable results for users and advertisers, and the changes we are making to our go-to-market organization are the right ones for the business long term.”
— Julia Brau Donnelly, Q1 FY2026
“Overall, large retailers remained a headwind to growth, but AI-driven platform improvements, including bidding optimizations we delivered for these advertisers, began to offset some of this headwind later in the quarter.”
— Julia Brau Donnelly, Q1 FY2026
“We are seeing strong early results there, including our efforts to link our AI bidding systems directly to advertisers' measurement sources of truth, and we plan to scale that pilot to additional large advertisers later this year.”
— Julia Brau Donnelly, Q1 FY2026
“What he is focused on first is bringing more accountability, more consistency, more operational rigor, and AI tooling to how we go to market.”
— William J. Ready, Q1 FY2026
“Some of the near-term changes I mentioned in my prepared remarks are already underway, including leadership changes across parts of the international and go-to-market organization, accelerating adoption of internal AI tools, and sharpening accountability across the sales force.”
— William J. Ready, Q1 FY2026
“For the first time, we have a product built to serve smaller advertisers that do not have the time, resources, or expertise to manage campaigns across multiple platforms, and we are only about a year into that journey, which we expect to be a multiyear cycle just as it was for the larger platforms when they deployed their AI-driven automation suites.”
— William J. Ready, Q1 FY2026
“We are adding more functionality across bidding, targeting, creative, and measurement over time, and a lot of that leverages our in-house capabilities and taste graph—things that we think we are really uniquely positioned to do and are demonstrating.”
— William J. Ready, Q1 FY2026
“Obviously nobody can perfectly predict the future, but we are actually several years into a massive AI adoption cycle, and that means we can really learn a lot from what people are already doing.”
— William J. Ready, Q1 FY2026
“It is important to note that at the same time chatbots have grown in popularity over the last few years, we have put up ten straight quarters of double-digit user growth and deepening engagement per user.”
— William J. Ready, Q1 FY2026
“Users, including Gen Z, are engaging with chatbots and Pinterest, Inc. at the same time for very different things.”
— William J. Ready, Q1 FY2026
“We surface relevant, personalized recommendations before the user even knows how to ask for what they want, and we connect that to real products that they can act on.”
— William J. Ready, Q1 FY2026
“We are solving the “I will know it when I see it” problem, which is such a significant component of so many consumer shopping journeys.”
— William J. Ready, Q1 FY2026
“On agentic commerce more broadly, you have also seen meaningful strategic pivots from some of the platforms that were most aggressively pursuing that space.”
— William J. Ready, Q1 FY2026
“We see ourselves as having effectively turned Pinterest, Inc. into an AI-powered shopping assistant that operates in a primarily visual manner, which is consistent with large portions of how people actually shop.”
— William J. Ready, Q1 FY2026
“Longer term, at the heart of our engagement strength is how we continue to leverage AI to drive better personalization and relevance.”
— William J. Ready, Q1 FY2026
“Our ongoing improvements to the platform, including the launches we highlighted this quarter across search ranking, content recommendations, and creative generation, are all pointing in the same direction, which is a more relevant and personalized experience that gives users more reasons to come back and anticipates what they are looking for next—all built off of our proprietary signals and unique curation behavior.”
— William J. Ready, Q1 FY2026
“That curation behavior that occurs on Pinterest, Inc., which we see as completely unique in the Western world, gives us a highly differentiated signal that we can use to train AI in ways that others without that signal cannot.”
— William J. Ready, Q1 FY2026
“We launched Pinterest, Inc. Assistant in beta in Q4 of last year.”
— William J. Ready, Q1 FY2026
“Over the past couple of months, we have materially advanced capabilities of the underlying model powering the Pinterest, Inc. Assistant, due to both advancements in the underlying open-source model as well as our ability to post-train that model with our unique data and integrate it into our suite of in-house models.”
— William J. Ready, Q1 FY2026
“Last, on models across the industry, the industry is converging on a view that we reached at Pinterest, Inc. relatively early on: the unit economics of relying on large proprietary third-party LLMs do not make sense for many use cases, as companies end up paying a significant premium for what might be an overengineered generalized capability that is not necessarily optimized for company-specific problems.”
— William J. Ready, Q1 FY2026
“Our approach has been deliberate from the start.”
— William J. Ready, Q1 FY2026
“We build compact, fit-for-purpose models trained on our proprietary data for our most unique and core use cases such as visual understanding, and we have seen these consistently produce better results at far lower cost for the majority of what our product does.”
— William J. Ready, Q1 FY2026
“For the more generalized LLM capabilities, we use suitable open-source models running in our own cloud environment within our infrastructure when they are the right tool, and then we post-train them on our own proprietary data.”
— William J. Ready, Q1 FY2026
“As we think about advancing our assistant, pairing our fit-for-purpose in-house models that are great at understanding and driving commerciality and recommendations with some basic LLM capabilities—and then post-training that in the places that can be helpful to the user—we think that unique combination can really help a lot and we can do differentiated things.”
— William J. Ready, Q1 FY2026
“That is one tangible example we talked about on the call of how we can use our data on top of algorithms to get even better outcomes and is part of what we are doing with AI models generally—both what we build in-house and where we retrain open-source models.”
— William J. Ready, Q1 FY2026
“AI is changing how people discover, how they form intent, narrow choices and move from inspiration to action.”
— William Ready, Q4 FY2025
“We've taken Pinterest from a platform with declining users into a growing AI-powered visual-first shopping assistant and search destination that has now put up 10 straight quarters of record high users.”
— William Ready, Q4 FY2025
“User growth accelerated in the second half as we continue to introduce AI-led features for both users and advertisers.”
— William Ready, Q4 FY2025
“Second, keep AI at the core of everything we do from highly personalized user experiences and new features like Pinterest Assistant to the advertiser experience through Pinterest Performance+ and to optimizing our own internal operations.”
— William Ready, Q4 FY2025
“Engagement is growing because our unique curation signal and taste graph, combined with cutting-edge AI has improved relevance significantly and made our surfaces much more actionable.”
— William Ready, Q4 FY2025
“Users open Pinterest to a personalized visual feed that starts their shopping journey without having to enter a prompt, bringing the Promise of Agentic commerce to life.”
— William Ready, Q4 FY2025
“In many ways, AI is following the same pattern cloud computing did over a decade ago.”
— William Ready, Q4 FY2025
“The winners will be the companies that combine those capabilities with truly differentiated data and solve problems in unique ways for users and customers.”
— William Ready, Q4 FY2025
“In 2025, our taste graph grew by nearly 40% as users make more associations across pins, products, boards, retailers and brands.”
— William Ready, Q4 FY2025
“Not only do we have differentiated signals, we're also leveraging AI in a highly capital-efficient manner.”
— William Ready, Q4 FY2025
“As a result, we use a combination of AI models, including our own proprietary fit-for-purpose foundation models, leading third-party proprietary models and increasingly open source models that we fine-tune on our unique signal.”
— William Ready, Q4 FY2025
“In 2025, we introduced OmniSage, our core AI model trained on our taste graph to turn those associations into a single high-value recommendation signal used to retrieve and rank content.”
— William Ready, Q4 FY2025
“The application of OmniSage drove a 450 basis point lift in site-wide saves.”
— William Ready, Q4 FY2025
“Additionally, in a continuation of our work to increase context windows and bring a user's full history across all major surfaces on Pinterest, we developed a proprietary foundation ranking model called PinFM.”
— William Ready, Q4 FY2025
“This model distills lifetime user actions into the recommendations on the home feed and related pens, driving personalization in nearly every impression our users see.”
— William Ready, Q4 FY2025
“This launch brought meaningful site-wide engagement gains, including a 240 basis point increase in saves across the platform.”
— William Ready, Q4 FY2025
“Finally, as open source models have made tremendous strides in performance, we developed a model framework called Navigator One, which allows us to leverage visual embeddings built on our taste graph and fine-tune open source models to power our newest AI-driven experiences.”
— William Ready, Q4 FY2025
“This framework reduces latency and delivers approximately 90% reduction in cost versus utilizing a leading third-party proprietary model.”
— William Ready, Q4 FY2025
“A great example of this is Pinterest Assistant, which we launched in beta in Q4.”
— William Ready, Q4 FY2025
“Pinterest Assistant is a voice-activated, visual-first conversational assistant that will leverage Navigator One to expand our multimodal discovery capabilities and seamlessly flow between images, voice and text.”
— William Ready, Q4 FY2025
“Compared with traditional text-based search, users are asking a significantly higher share of commercially oriented questions, about 25 percentage points more when using Pinterest Assistant.”
— William Ready, Q4 FY2025
“Lastly, AI is at the core of how we are improving efficiencies internally as roughly 50% of our new code is AI generated.”
— William Ready, Q4 FY2025
“More diverse advertiser demand allows our models to serve more relevant, highly personalized ads to users.”
— William Ready, Q4 FY2025
“Late last year, we began to pilot integrations with a few of our most sophisticated advertisers proprietary in-house measurement systems to help us optimize bids to drive more of the outcomes those advertisers value.”
— William Ready, Q4 FY2025
“So far, this pilot has delivered promising results with one advertiser increasing its bids on Pinterest by more than 30%, reflecting the higher value it was seeing from the platform under this new value-based optimization approach.”
— William Ready, Q4 FY2025
“These integrations will allow automated 2-way data transfer, so we can continuously train and optimize our bidding models to reflect advertisers' highest valued outcomes and thus show up more favorably in their measurement systems.”
— William Ready, Q4 FY2025
“In initial testing, advertisers saw new customer conversions increase by an average of 64% in campaigns where new customer acquisition was enabled compared to control campaigns without it.”
— William Ready, Q4 FY2025
“We're proud to lead the way by tuning our AI for positivity and giving our users more agency and choice over their experience.”
— William Ready, Q4 FY2025
“The increase was driven by headcount investments in sales and marketing and R&D as we continue to invest in AI initiatives and grow our sales force.”
— Julia Donnelly, Q4 FY2025
“In Q1, within non-GAAP operating expense, we will focus our investments on our sales transformation and additional R&D hiring to support our AI efforts.”
— Julia Donnelly, Q4 FY2025
“In 2026, we're making deliberate investments in high ROI areas such as GPU capacity to enable key AI initiatives.”
— Julia Donnelly, Q4 FY2025
“These investments will allow us to train and serve visual foundation models and our conversation models that advance our capabilities in multimodal search, discovery as well as Pinterest Assistant.”
— Julia Donnelly, Q4 FY2025
“In addition, we will continue to build more powerful AI models that are enhancing full funnel ROAS for our advertisers and adds relevance for our users.”
— Julia Donnelly, Q4 FY2025
“The result of these offsetting dynamics is that we expect adjusted EBITDA margins to be roughly in line with 2025.”
— Julia Donnelly, Q4 FY2025
“Now we expect to reinvest roughly half of those OpEx savings primarily in our sales transformation and in AI talent.”
— Julia Donnelly, Q4 FY2025
“We anticipate adjusted EBITDA margins, as I said on the call, to be kind of roughly in line with 2025, excluding the approximately 100 basis point drag from the tvScientific acquisition, which results in sort of 29% for full year 2026 overall.”
— Julia Donnelly, Q4 FY2025
“But we believe this will drive further improvements to advertiser performance and therefore, advertiser budgets and continued user and engagement growth.”
— Julia Donnelly, Q4 FY2025
“SMBs who are adopting Performance Plus campaigns to automate and simplify campaign setup with AI are seeing stronger performance and are spending more on our platform.”
— William Ready, Q4 FY2025
“So as we noted last quarter, we see a 12% higher monthly revenue growth rate with these managed SMB advertisers versus nonadopters.”
— William Ready, Q4 FY2025
“Even as AI advances, we talked about how we're able to use low-cost open source AI in our own internal proprietary models, train that against that data and then get very different results.”
— William Ready, Q4 FY2025
“I shared on prior calls, that our latest multimodal visual search models outperform leading proprietary off-the-shelf models by 34 percentage points on the relevancy of shopping recommendations.”
— William Ready, Q4 FY2025
“Our visual search discovery and personalization means that users are instantly met with relevant products that they're interested in when they open up the Pinterest app.”
— William Ready, Q4 FY2025
“We're helping them complete those commercial journeys without having to type in a single prompt.”
— William Ready, Q4 FY2025
“In essence, we're helping our users know what to buy before they know what to ask for, which has historically been one of the biggest problems in search is that people don't have the words to describe what it is they're looking for.”
— William Ready, Q4 FY2025
ANALYST QUESTIONS ON AI
Q (Q1 FY2026, Colin Alan Sebastian): Performance Plus now running at approximately 30% of lower-funnel revenue. What adoption trends are you seeing within the mid-market and SMB segments? And related to that, given that Performance Plus adopters are growing their spend at, I think, twice the rate of non-adopters, how are you leveraging tools like Canvas and PinRack to lower those barriers for smaller advertisers?
A: William J. Ready: Early adoption is encouraging. The 30% of our lower-funnel revenue that is now running through Performance Plus campaigns—we feel good about that, but obviously that is still early in the journey... We are adding more functionality across bidding, targeting, creative, and measurement over time, and a lot of that leverages our in-house capabilities and taste graph—things that we think we are really uniquely positioned to do and are demonstrating.
Q (Q1 FY2026, Jason Helfstein): How are you viewing the impact from chatbots with respect to the competitive landscape and emerging visual discovery?
A: William J. Ready: At the same time chatbots have grown in popularity over the last few years, we have put up ten straight quarters of double-digit user growth and deepening engagement per user... We surface relevant, personalized recommendations before the user even knows how to ask for what they want, and we connect that to real products that they can act on. We are solving the “I will know it when I see it” problem... Pinterest, Inc. is a specialized platform, and that is a position of strength.
Q (Q1 FY2026, Justin Patterson): What do you see as the core levers to continue doing [deepening engagement]? Given UCAN is a more established market, how much more runway do you have to drive further engagement growth here?
A: William J. Ready: We see ourselves as having effectively turned Pinterest, Inc. into an AI-powered shopping assistant that operates in a primarily visual manner... Longer term, at the heart of our engagement strength is how we continue to leverage AI to drive better personalization and relevance. Our ongoing improvements to the platform, including the launches we highlighted this quarter across search ranking, content recommendations, and creative generation, are all pointing in the same direction... That curation behavior that occurs on Pinterest, Inc., which we see as completely unique in the Western world, gives us a highly differentiated signal that we can use to train AI in ways that others without that signal cannot.
Q (Q1 FY2026, Ronald Victor Josey): Talk to us about how retailers are preparing for this going forward. As you look out maybe one to three years and we hear about the personal assistant on Pins, how do you envision that future going forward?
A: William J. Ready: We launched Pinterest, Inc. Assistant in beta in Q4 of last year... Over the past couple of months, we have materially advanced capabilities of the underlying model powering the Pinterest, Inc. Assistant, due to both advancements in the underlying open-source model as well as our ability to post-train that model with our unique data and integrate it into our suite of in-house models... We build compact, fit-for-purpose models trained on our proprietary data for our most unique and core use cases such as visual understanding... For the more generalized LLM capabilities, we use suitable open-source models running in our own cloud environment within our infrastructure when they are the right tool, and then we post-train them on our own proprietary data.
Q (Q1 FY2026, Eric James Sheridan): What signal investors should be taking in terms of what [Lee Brown's hiring] means for your go-to-market strategy...?
A: William J. Ready: What he is focused on first is bringing more accountability, more consistency, more operational rigor, and AI tooling to how we go to market... Some of the near-term changes I mentioned in my prepared remarks are already underway, including leadership changes across parts of the international and go-to-market organization, accelerating adoption of internal AI tools, and sharpening accountability across the sales force.
Q (Q4 FY2025, Eric Sheridan): When you think about ChatGPT and they're launching their own ad product and you have a lot of ambition for growth across the industry at the same time that the industry is moving towards more automation and more AI and machine learning. Can you bring together your vision for how you see Pinterest broadly fitting into this increasingly competitive landscape for digital advertising budget dollars?
A: William Ready: We have a unique curation signal. We have a differentiated full funnel platform... Even as AI advances, we talked about how we're able to use low-cost open source AI in our own internal proprietary models, train that against that data and then get very different results. I shared on prior calls, that our latest multimodal visual search models outperform leading proprietary off-the-shelf models by 34 percentage points on the relevancy of shopping recommendations.
Q (Q4 FY2025, Justin Patterson): You mentioned earlier that Pinterest visual feed brings the Promise of Agentic commerce to life without having to enter prompts. Could you expand some more on just what Agentic commerce means for Pinterest and the steps to get there?
A: William Ready: The Promise of Agentic is one where users trust AI to help them along a commercial journey to remove friction and to find products they love, all without the user having to do as much of the work. That's exactly where Pinterest has been leaning in. Our visual search discovery and personalization means that users are instantly met with relevant products that they're interested in when they open up the Pinterest app. We're helping them complete those commercial journeys without having to type in a single prompt... In Q4, we accelerated our product even further, introducing Pinterest Assistant, which adds voice as a new modality.
Q (Q4 FY2025, Kenneth Gawrelski): How much tech investment beyond just kind of sales and go-to-market, but more tech investment might be necessary to broaden that advertiser base and broaden and deepen that advertiser base?
A: William Ready: SMBs who are adopting Performance Plus campaigns to automate and simplify campaign setup with AI are seeing stronger performance and are spending more on our platform. So as we noted last quarter, we see a 12% higher monthly revenue growth rate with these managed SMB advertisers versus nonadopters. So the ongoing improvements we're making to Performance Plus around measurement and attribution will be particularly important for this group...
Q (Q4 FY2025, Colin Sebastian): Could you maybe walk through in a little more detail the puts and takes on the adjusted EBITDA outlook for the year just as we move through the year and then you balance some of those -- the impacts from some of those various factors?
A: Julia Donnelly: We're intentionally investing in cost of revenue, specifically in GPU capacity to enable key AI initiatives... But we believe this will drive further improvements to advertiser performance and therefore, advertiser budgets and continued user and engagement growth. So we expect this cost of revenue investment to be approximately 100 basis points in 2026... Now we expect to reinvest roughly half of those OpEx savings primarily in our sales transformation and in AI talent.