The short version
We set out to build a straight lease-versus-buy ROI table for restaurant delivery robots, using published Robots-as-a-Service rates. We couldn’t, and the reason is the story. Every price we re-checked on 2026-08-19:
- The same robot is listed at two or three different monthly rates on the same seller’s website — the Bear Servi Plus appears at $479/mo, $293/mo, and $1,505/mo across three parts of RobotLAB’s site. That’s a 5.1× spread.
- Purchase prices contradict themselves on a single page. The Keenon T10’s page header says “Starting at $18,400” directly above a line reading “$13,500 purchase or $542/mo RaaS.”
- The one number that would make buying comparable — an ongoing service-plan price — isn’t really published. The widely-cited “$199/mo” plan appears only in a page’s HTML metadata; the actual service-plans page lists four tiers with no prices at all and different terms.
- Two of the three manufacturers publish no prices whatsoever. Keenon lists none. Pudu lists none, and its own service-plan page contains zero dollar figures.
- One useful pattern did hold. Across RobotLAB’s store-listing badges, RaaS lands at a consistent 3.36–3.42% of purchase price per month — meaning 36 months of leasing runs about 121–123% of buying outright.
If you take one thing from this page: the published number is not the number. Get a written quote, and make the seller tell you which of their own figures applies.
The rate confusion, documented
RobotLAB is the most useful seller here precisely because it publishes monthly rates at all — most don’t. But it publishes them in up to four places per robot (a store-listing badge, a product-page hero, a spec table, and the page’s own description text), and those places disagree.
| Robot | Listing badge | Page hero | Spec table | Description text |
|---|---|---|---|---|
| Bear Servi | $339/mo · $9,990 | $9,990 | buy $11,990 | “From $11,990” |
| Bear Servi Plus | $479/mo · $13,990 | $13,990 | $293/mo | “RaaS from $1,505” |
| Pudu BellaBot | $489/mo · $14,500 | $14,500 | — | “from $15,900, RaaS from $2,430/mo” |
| BellaBot Pro | $399/mo · $16,000 | $16,000 | $335/mo | — |
| Pudu KettyBot Pro | $479/mo · $14,000 | $14,000 | $395/mo | “Buy from $9,500, RaaS $395/mo” |
| Keenon DINERBOT T10 | $619/mo · $18,400 | $18,400 | buy $13,500, $542/mo | “$13,500 purchase or $542/mo RaaS” |
The Servi product page states it plainly in its own body copy: “This page reflects RobotLAB’s purchase price of $11,990” — while the header of that same page reads $9,990.
It isn’t only one seller. Robots International’s DINERBOT T10 page carries $9,995 in its cart button and its machine-readable product data, while the same page’s description and FAQ say the T10 costs $23,000 — and cite the T9 as “$9,800” when that site’s own T9 list price is $7,995. Homery lists two KettyBot Pro entries with near-identical titles at $11,000 and $14,000.
What this means for you: a quote is not a market price, and a website number is not a quote. When three published figures exist, the seller will honour whichever one you negotiated in writing.
The only self-consistent pattern we found
Strip out the contradictions and look only at RobotLAB’s store-listing badges, which are internally coherent. The lease rate is a near-constant fraction of purchase price:
| Robot | Buy | RaaS/mo | Monthly % of purchase | 36 months | vs. buying |
|---|---|---|---|---|---|
| Bear Servi | $9,990 | $339 | 3.39% | $12,204 | 122% |
| Bear Servi Plus | $13,990 | $479 | 3.42% | $17,244 | 123% |
| Pudu BellaBot | $14,500 | $489 | 3.37% | $17,604 | 121% |
| Pudu KettyBot Pro | $14,000 | $479 | 3.42% | $17,244 | 123% |
| Keenon T10 | $18,400 | $619 | 3.36% | $22,284 | 121% |
| BellaBot Pro | $16,000 | $399 | 2.49% | $14,364 | 90% |
Five of six robots cluster tightly: 36 months of RaaS costs roughly one-fifth more than buying the same machine. That is a defensible rule of thumb — and, to be clear, it’s a pattern we derived from one seller’s listing badges, not an official rate card.
The BellaBot Pro is the outlier at 90%, which is either a genuinely better deal or a stale badge. It’s worth asking about specifically.
The 20% premium is not the whole comparison, though, because leasing bundles things buying doesn’t. Which brings us to the number that isn’t published.
The buy side is not $0/month
A lease at ~121% of purchase price looks expensive until you ask what a bought robot costs to keep running. RobotLAB’s RaaS is described as all-inclusive: hardware, site survey, on-site deployment by engineers, staff training, technical support, maintenance, and software updates. A purchased robot needs those too.
Here’s where it falls apart. The commonly repeated figure is a “$199/mo per unit” Standard service
plan covering 24/7 support, parts and labour, a loaner if downtime exceeds 48 hours, quarterly
health checks and an annual on-site visit. We could not find that plan on the service-plans
page. The $199 figure appears in that page’s title tag and meta description only — nowhere in
the visible page. The visible page lists four tiers — Basic, Silver, Gold, Platinum — with no
prices, no tier called “Standard,” and terms that don’t match: a 72-hour replacement swap
rather than a 48-hour loaner, and one preventive maintenance visit per year rather than
quarterly.
Pudu’s own service programme has the same gap. PUDU Care publishes real substance — the Standard package covers unlimited on-site repair with no cap on visit count, four annual preventive maintenance visits, one installation, and spare parts; Basic covers remote support, one installation and two on-site maintenance visits; both promise a 2-hour service response with issues “resolved in as fast as 2 business days” — and both are US and Canada only. What the page does not contain, anywhere, is a dollar figure.
So the honest statement is: nobody publishes the recurring cost of owning one of these robots. Until you have that number in writing, you cannot complete the comparison — a service plan anywhere near $199/mo would consume most of the ~20% that leasing costs extra.
Other terms worth pinning down: the 36-month standard term with convert-to-purchase after month 12 appears exactly once across RobotLAB’s site, in an FAQ block — it’s absent from the financing page and from every product page. The financing page itself publishes no RaaS rates at all (“Ask your specialist for current RaaS pricing on the robot you need”), quotes a different floor than the store page does, and offers a $1 buyout lease structure through SBA-approved partners Ascentium and Conners Financial. Its tax language is appropriately hedged: “Lease payments might be tax-deductible as operating expenses.”
The labour side of the ROI — and two caveats that break most robot math
The usual pitch is that a robot replaces or supplements a server. The current federal wage data, from the May 2025 Occupational Employment and Wage Statistics release:
| Occupation | Employment | Mean hourly | Mean annual | Median annual |
|---|---|---|---|---|
| Waiters and waitresses (35-3031) | 2,270,910 | $19.26 | $40,060 | $35,230 |
| Dining room and cafeteria attendants (35-9011) | 542,750 | $16.46 | $34,230 | $33,980 |
Two caveats that most robot ROI marketing ignores, both stated by BLS itself:
- Tips are included in those figures. The wage definition explicitly includes tips. So $40,060 is not what a waiter costs an employer — a large share is money guests hand over directly. Any claim that a robot “replaces $40,000 of labour cost” is wrong.
- The annual figure is arithmetic, not observation. BLS computes annual wages by assuming a “full-time, year-round schedule of 2,080 hours” ($19.26 × 2,080 = $40,060). Restaurant service work is heavily part-time, so treat it as a hypothetical full-timer, not a typical one.
Do the comparison with your own payroll cost per hour, not the headline number. As a sense of scale: at a $479/mo badge rate, a Servi Plus lease costs about what 25 hours of work costs at the $19.26 mean — under six hours a week. That framing is honest only if you substitute your real loaded labour rate, tips excluded.
Turnover is the one labour statistic that’s cleanly citable. In the June 2026 JOLTS data (released 2026-08-04), the quits rate for accommodation and food services was 4.5% — 638,000 people — against 2.0% for total nonfarm. Restaurants shed staff at roughly 2.25× the all-industry rate. Note that’s a monthly rate, and it’s down from 5.5% a year earlier. If your robot’s value is “the shift runs when you’re two people short,” that’s the number that supports it — better than any productivity claim we found from a vendor.
And be sceptical of those. One seller’s page asserts that two extra table turns per evening at $150 per table is “roughly $100,000 in additional annual revenue” — with no methodology and no source. Pudu’s own US page advertises “400 Deliveries Per Day” and “$10/Day for a Usage Period of 5 Years.” Note what that second claim is: a five-year total-cost-of-ownership figure, not a lease rate — and depending on whether you count 365 or 260 days a year it implies $18,250 or $13,000, which the page never specifies. Pudu also runs a two-week BellaBot trial for $299 in Houston, Dallas, San Antonio and Austin (shipping extra beyond 100 miles), routed through its retailer American Robotech. At $299 for two weeks in your actual dining room, that trial is a better source of ROI data than any table on any website, including this one.
Tax: the one place buying has a structural edge
For tax years beginning in 2026, IRS Revenue Procedure 2025-32 sets the Section 179 expensing limit at $2,560,000, phasing out once you place more than $4,090,000 of qualifying property in service. Practically no restaurant will approach either ceiling, which means a purchased robot can generally be expensed in full in year one rather than depreciated.
The catch for leases is in IRS Publication 946: “To qualify for the section 179 deduction, your property must have been acquired by purchase,” and if you lease property from someone else you “generally cannot depreciate its cost because you do not retain the incidents of ownership.” Lease payments are instead deducted as rent — but Publication 334 adds a condition that matters here: “If you have or will receive equity in or title to the property, you can’t deduct the rent.”
That language sits awkwardly against a $1 buyout lease, which by design conveys title. We are not going to resolve how any specific agreement gets treated — that depends on the contract, and it’s a question for your CPA. Ask it before you sign, because the answer can move the lease-versus-buy comparison by more than the ~20% premium we calculated above.
How to actually decide
The published data won’t decide this for you. These four questions will:
- “Which of your published prices applies to me?” Name the specific figures — the badge, the hero, the spec table. Get the answer in writing.
- “What does the service plan cost per month if I buy?” Without it you can’t compare. Get the tier name, the price, the response-time SLA, and what happens when the robot is down.
- “What are the exact lease terms?” Length, buyout structure, whether convert-to-purchase after month 12 is real and in the contract, and what happens if you close or relocate.
- “Can I trial it first?” A paid trial in your own room beats every projection.
Lease if cash is tight, you want service bundled and accountable to one throat to choke, or you’re not certain the robot fits your floor plan. Buy if you have the capital, can get the service plan priced, and can use the Section 179 deduction — you’ll likely save on the order of 20% over three years, assuming the ongoing service cost doesn’t eat it.
One last data point on how unreliable published pricing is here: across every seller we checked, the only price that replicated independently was the Pudu BellaBot Pro at $16,000, listed identically by RobotLAB and by Homery. One price in the entire category agreed with itself.