ThorData Proxy and Web Data Infrastructure Review

OpenSourcesAI verdict

We have not run production traffic through ThorData, so read this as a desk review of the published product rather than a benchmark. On paper the offer is legible: four proxy types, a SERP and scraper API layered on top, pay-as-you-go billing. The number to be careful with is the headline price. Residential starts at $2.00/GB and only reaches the advertised $0.65/GB once you are buying 5TB a month. Most teams never get near 5TB. Price your real monthly volume against the tier you would actually land on, compare that against Bright Data and an official API if one exists, and the decision usually makes itself.

Best for

Teams whose collection is genuinely blocked by IP reputation, geo-targeting or rate limits, and who have already settled that the collection itself is lawful. It is a poor fit for scraping one site occasionally, and a bad fit for anything an official API already serves.

Why use it

Buy proxies when the thing standing between you and the data is the IP address. If your scraper is breaking on layout changes, malformed markup or session state, a proxy pool fixes none of that, and you will pay per gigabyte to keep failing. Diagnose which wall you have hit before shopping.

Product overview as of August 2026

Checked against thordata.com on 11 August 2026. Four proxy types are sold: residential, static ISP, mobile and datacenter. Six data products sit above them: SERP API, Web Scraper API, Web Unlocker, Scraping Browser, Video Data Scraper and Video Datasets. The residential pool is advertised at 100M+ IPs across 190+ countries, with per-country counts published on the homepage (3,876,096 US, 1,850,036 UK, 1,267,993 Germany). Uptime is claimed at 99.9% for residential and 99.99% overall, with sub-0.5s response. Those are vendor figures. No one outside the company can audit a proxy pool count or a success rate, and we have not tried to.

Where it fits

• Collection layer. This is the only layer ThorData occupies: getting a request to land from the right place, often enough to be useful.
• It does not parse, schedule, deduplicate or store. Those stay yours, or come from a different tool.
• The scraping APIs blur that line, since they return structured results for common targets. Check coverage for your specific sites before assuming they replace a parser.
• If you are feeding a RAG or training pipeline, the proxy sits upstream of every data-quality problem you will have later. Garbage collected at scale is still garbage.

Common AI and business use cases

• Price and availability monitoring across regions where one IP gets throttled within minutes.
• SERP and AI-answer visibility tracking at a volume manual checks cannot cover.
• Localization QA: seeing what a page actually renders as from Tokyo, Berlin or Sao Paulo.
• Building public corpora for RAG or evaluation sets, where the legal basis is already settled.
• Anything you currently run from a single office IP that keeps hitting rate limits.

Evaluation checklist

• Is this collection lawful and permitted for your specific target and jurisdiction? Settle that before pricing anything.
• Does an official API exist? It will be cheaper, more stable and far less contentious than any proxy.
• What volume will you actually use per month? Read the price at that tier, not at the advertised floor.
• Residential, ISP, mobile or datacenter? Datacenter is cheapest and gets blocked first. Residential costs more per gigabyte and more per failed request.
• How are failed and retried requests billed? On per-GB pricing a blocked response still moves bytes.
• Do the scraping APIs cover your targets, or are you writing parsers and paying for proxies?

Security and admin notes

• Collect public data only. A proxy changes what is possible, not what is permitted.
• Do not route past authentication, paywalls or consent walls.
• Personal data collected at scale pulls in GDPR and CCPA obligations even when each record was publicly visible.
• Keep proxy credentials and API keys in a secrets manager. Leaked proxy credentials get resold and the traffic gets billed to you.
• Log what you collected, from where, and on what basis. You will need that record before you want it.

Pricing notes

Checked 11 August 2026. Residential is pay-as-you-go on a sliding per-GB scale: $2.00/GB at 1GB, $1.50 at 50GB, $1.00 at 150GB, $0.80 at 350GB, and $0.65 at 5TB, which is a sale rate against a $1.05 list price. Mobile runs far higher, $5.00/GB down to $2.20/GB at 500GB. Static ISP and datacenter are per-IP at $0.75, on sale from $1.50. APIs are per thousand: Web Scraper from $0.50/1K results, SERP from $0.70/1K responses, Web Unlocker from $1.00/1K. Two things to watch. Several headline rates are sale rates sitting next to list rates, so confirm which one applies at checkout. And because per-GB pricing is tiered by monthly volume, your effective rate depends entirely on committed spend, which makes lowest-tier comparisons between vendors close to meaningless.

Tradeoffs

What you are renting is IP reputation, and that is a reasonable thing to rent. Running your own pool is miserable and rarely worth it. What you take on instead is a metered dependency where cost scales with failure: every block, retry and oversized response is billable traffic. The vendor that looks cheapest per gigabyte is not always cheapest per successful record, and cost per successful record is the only figure worth comparing.

Pros

• Four proxy types on one account, so a job can move from datacenter to residential without changing vendors.
• Per-IP pricing on ISP and datacenter is predictable in a way per-GB billing is not.
• Per-thousand API pricing makes cost per record easy to estimate before you commit.
• Pay-as-you-go with no visible minimum, so a small pilot is cheap to run.

Cons

• Pool sizes, uptime and response times are vendor-published and not independently verifiable.
• The advertised $0.65/GB requires a 5TB month. Realistic volumes land two to three times higher per gigabyte.
• Sale pricing sits alongside list pricing across the site, so the figure you compared may not be the figure you pay.
• Legal exposure for what you collect stays entirely with you.
• Not open source, not local, and nothing here runs on hardware you control.

Alternatives

• Bright Data is the incumbent, with larger pools and more compliance tooling, priced accordingly.
• Oxylabs and Smartproxy sit in the same bracket. Get a quote for your actual volume, since list pricing rarely survives contact with a sales team.
• Browse AI is the better answer if you want monitoring and extraction without writing code.
• For one site at low frequency, a small VPS and polite rate limiting will do the job.
• An official API, where one exists, beats every option above.

Recommended workflow

• Write down the exact data, the source, and why collecting it is permitted. If you cannot finish that sentence, stop here.
• Check for an official API before anything else.
• Estimate monthly volume in gigabytes or requests, then read the price at that tier.
• Start on datacenter or ISP. Move to residential only once you can show the cheaper tier is being blocked.
• Run a metered pilot and measure cost per successful record, retries included.
• Add backoff, caching and logging before scaling. Caching usually cuts the bill more than any volume discount.

FAQ

Related OpenSourcesAI pages