TaskUs: 3.8x EBITDA, but the Replacement Engine Is Unproven

At roughly $6 per share, TaskUs has the shape of a classic broken-stock opportunity.¹ Meta is automating Trust & Safety work, the shares are beaten down, and the rest of the business is still growing. The raw chart overstates the loss because TaskUs paid a $3.65 special dividend in March, but adding that distribution to the current price still leaves cumulative value about 48% below the 52-week high. Using 91.6 million economic shares, $152 million of cash and $500 million of debt, the enterprise value is about $892 million. That is roughly 3.8x the adjusted EBITDA implied by TaskUs’s FY2026 guidance.²

The valuation is real. So is the growth. The problem is that TaskUs has not shown that the growth can durably replace the revenue and profit being automated.

Q1 did not run that test. Meta revenue still grew approximately 1%, Trust & Safety still grew 4.7%, and every service line remained positive. Management expects the Meta-led Trust & Safety decline to begin in Q2. Q1 proved that TaskUs has offsets before the contraction. It did not prove that those offsets can absorb it.³

That leaves TaskUs highly interesting and still too difficult to underwrite. Three variables determine whether 3.8x EBITDA is a bargain or a misleading multiple: how long the replacement growth lasts, what margin it earns, and which human workflows AI removes next. TaskUs does not disclose enough to answer any of them.

At a glance

Item	Snapshot
Listing	Nasdaq · TASK
Price	$5.98 · July 15, 2026 close
Market cap	Approximately $548M
Total-return performance	1M +11% · 6M −21% · 1Y −47%
Broad valuation	~3.8x FY26 guided EV/adjusted EBITDA · ~4.5x NTM P/E · ~4.7x market cap/NTM FCF

Performance uses dividend-adjusted prices; the unadjusted six-month share-price change is approximately −48% because TaskUs paid a $3.65 special dividend in March.⁴

Financial snapshot

$M, except growth and margin	2021A	2022A	2023A	2024A	2025A	2026E	2027E
Revenue	760.7	960.5	924.4	995.0	1,183.5	1,233	1,306
Growth	59.1%	26.3%	(3.8%)	7.6%	19.0%	4.2%	6.0%
Adjusted EBITDA margin	24.7%	23.2%	23.9%	21.1%	21.0%	18.8%	19.6%
Operating cash flow	(32.7)	147.1	143.7	138.9	137.2	—	—
Free cash flow	(92.0)	103.3	112.7	99.8	73.7	—	—

Historical revenue, adjusted EBITDA margin and cash flow reconcile to TaskUs’s annual filings. Free cash flow is TaskUs’s definition: operating cash flow less purchases of property and equipment.⁵

The business is already part disruptor, part disrupted

TaskUs provides outsourced digital operations for technology and consumer companies. Digital Customer Experience handles support, sales and other customer workflows. Trust & Safety combines content moderation with Financial Crime & Compliance. AI Services supplies data, evaluation, safety testing and operational support for foundation models, autonomous vehicles and robotics.

The business performed well in 2025. Revenue rose 19.0% to $1.184 billion, adjusted EBITDA rose 18.7% to $249.1 million, and adjusted EBITDA margin held near 21%. AI Services supplied $214.2 million of revenue and grew 58.6%.⁶

That history matters because the bear case cannot simply be “AI destroys BPO.” TaskUs is winning new work from AI deployment while AI removes older work. The investment question is whether the new work grows faster than human intensity falls across customer contacts, moderation decisions, model-development tasks and vehicle miles.

The public data show both forces. They do not show which one wins.

Clients 2–20 are growing, but this is not broad diversification

TaskUs highlighted two reassuring Q1 figures: revenue outside Meta grew 13.5%, and clients 2–20 grew well north of 20%.⁷ Those numbers are consistent with genuine wallet-share gains. They are also easy to overread.

Applying TaskUs’s rounded concentration disclosures to reported revenue produces the following directional bridge:

Client bucket	Q1 2025	Q1 2026	Implied growth
Meta	$72.2M	$73.5M	1.8%
Clients 2–20	$122.2M	$156.2M	27.8%
Outside the top 20	$83.3M	$76.6M	(8.1%)

The percentages are rounded and client ranks can change, so this is not a fixed-cohort analysis. The direction is still clear. TaskUs grew by expanding several large accounts, not because its entire client base accelerated.

Management said more than 75% of Q1 signings came from existing clients, approximately 70% of growth came from relationships older than one year, and more than 80% of AI Services growth came from existing clients, led by one long-term autonomous-vehicle customer.⁸

That can be a good business. A vendor can take wallet share while a client’s total outsourced workload shrinks, especially when the client consolidates suppliers. But it replaces one concentration with several others. It does not establish that clients 2–20 are immune to the same automation occurring at Meta.

Physical AI is the best offset—and still contains its own automation mechanism

The strongest evidence of genuinely new demand is in autonomous vehicles and robotics.

TaskUs’s Q1 presentation described new statements of work supporting a robotaxi provider’s multi-city expansion and providing training and remote operations to multiple robotics and autonomous-logistics developers.⁷ Its AV offering now includes remote intervention, stuck-vehicle analysis, roadside recovery, depot operations, rider support, emergency dispatch and law-enforcement coordination.⁹

This is not legacy content moderation with a new label. Robotaxi deployment creates real workflows that did not exist before. A stranded vehicle, collision or emergency can still require remote judgment or a person in the field. Physical-AI operations are the most credible multi-year replacement engine in the current portfolio.

The governing equation is nevertheless simple: deployed vehicles and miles multiplied by human interventions per mile. Fleet growth can outrun falling intervention rates for years. Better autonomy should eventually reduce the work required for each mile. TaskUs does not disclose physical-AI revenue, margin, client concentration, fleet exposure or intervention volumes, so investors cannot observe either side of that equation.⁸

Other AI Services work is less obviously durable. Revenue in the segment rose from $104.9 million in 2021 to $144.5 million in 2022, fell 8.9% to $131.7 million in 2023, barely grew in 2024, and then reached $214.2 million in 2025.⁶ That is a real business with a project-sensitive history, not a smooth recurring-revenue stream. TaskUs’s contracts generally lack committed volumes and can often be terminated for convenience.⁶

Some of the new work directly cannibalizes the old. The Q1 deck highlighted an AI Services engagement training a social-media company’s system to automate content moderation.⁷ Current AI revenue is therefore helping remove future Trust & Safety revenue. TaskUs also promotes agentic tools that automate customer interactions.⁷ Complex support, sales, regulated exceptions and safety work may persist longer, but TaskUs does not disclose enough workload mix to size them.

Revenue replacement is running ahead of profit replacement

Q1 revenue rose $28.5 million. Cost of services rose $26.6 million. Only $1.9 million of the increase remained after direct service costs, while adjusted EBITDA fell $0.7 million and margin declined from 21.3% to 19.1%.³

That does not prove AI Services has structurally poor margins. New sites, wage inflation, security costs, delivery geography and investment spending all affect the group result. TaskUs reports one operating segment and does not disclose margin by service line, workload, client or billing model.

It does prove that revenue replacement and economic replacement are different questions.

Management’s own guidance makes the distinction visible. At the midpoint, FY2026 revenue rises 3.5% to $1.225 billion. A 19% adjusted EBITDA margin implies approximately $232.8 million of adjusted EBITDA, down 6.6% from 2025.² The guide allows revenue replacement while group earnings shrink. It does not isolate replacement margins, but the bullish offset thesis has not yet translated into protected EBITDA.

The low multiple only protects the equity if the denominator remains reasonably stable. A bounded Meta contraction with continued ex-Meta growth and cost flexibility can leave the shares cheap. If automation spreads through Digital Customer Experience while replacement work earns lower margins, the current EBITDA figure is not a usable valuation anchor.

Why this remains a pass

The evidence resolves the easy versions of the debate.

TaskUs is not “only Meta”: management has identified similar Trust & Safety automation elsewhere, and TaskUs itself is helping clients automate moderation and customer support.⁷ The rest of the business is not simply waiting to disappear either: wallet-share gains are real, regulated exceptions remain labor-intensive, and physical-AI deployment is creating new operating work.

• Duration is unknown. AI Services has already contracted once, foundation-model work is project-sensitive, and physical-AI workload intensity is undisclosed.
• Margin is unknown. Q1 revenue grew while adjusted EBITDA fell, and TaskUs does not report replacement-work economics.
• The disruption boundary is unknown. TaskUs does not split simple from complex customer experience, content moderation from Financial Crime & Compliance, or physical operations from transitional data work.
• The growth is concentrated. Clients 2–20 expanded, while the revenue outside the top 20 appears to have declined.

None of these gaps proves that earnings will collapse. Together, they prevent a durable revenue and margin denominator from being estimated with useful precision.

What would change the answer

Q2 and Q3 should be judged against the actual contraction, not another ex-Meta growth headline.

TaskUs needs to show that total revenue and adjusted EBITDA dollars hold while Trust & Safety declines. Clients outside the top 20 need to stabilize. Physical-AI growth needs to broaden beyond one leading AV relationship. Margin should recover after the guided Q2 trough even if AI Services continues growing as a share of revenue.

Better disclosure could also make the case tractable: physical-AI workload economics, content-moderation versus Financial Crime & Compliance revenue, or evidence that outcome-priced agentic deployments let TaskUs retain the productivity gain.

Until then, TaskUs may be cheap enough for a client-specific impairment if margins hold. It is not demonstrably cheap enough for broad automation. Q1 showed growth before the contraction. Q2 and Q3 must show growth—and profit replacement—through it.

---

Evidence through July 15, 2026. Valuation uses a TaskUs share price of approximately $5.94, 91.6 million economic shares and balance-sheet figures at March 31, 2026. Client-bucket figures are Arc calculations using rounded company concentration disclosures and may be affected by changes in client rank.

Notes

1. TaskUs historical price lookup, accessed July 15, 2026. ↩
2. TaskUs Q1 FY2026 results and FY2026 outlook, May 6, 2026. ↩
3. TaskUs Q1 FY2026 Form 10-Q, filed May 7, 2026. ↩
4. Arc market data as of July 15, 2026. Market cap uses 91.6 million economic shares. Enterprise value for the FY2026 multiple uses $500.0 million of face debt and $152.3 million of cash at March 31, 2026; the NTM ratios use Arc’s provider definitions. ↩
5. TaskUs FY2022 Form 10-K, filed March 6, 2023; FY2023 Form 10-K, filed March 8, 2024; FY2024 Form 10-K, filed March 6, 2025; and FY2025 Form 10-K, filed March 5, 2026. ↩
6. TaskUs FY2025 Form 10-K, filed March 5, 2026. ↩
7. TaskUs Q1 FY2026 investor presentation, May 6, 2026. ↩
8. TaskUs Q1 FY2026 conference call and investor materials, May 6, 2026. ↩
9. TaskUs Autonomous Vehicle Operations, accessed July 15, 2026. ↩