European Alternatives to DeepSeek
Looking for a European alternative to DeepSeek? DeepSeek's services are operated by Hangzhou DeepSeek Artificial Intelligence Co., Ltd., whose terms are governed by the laws of the People's Republic of China (mainland) and whose privacy policy says personal data is collected, processed and stored in China. It prices its API per million tokens, with lower off-peak rates, and offers an OpenAI-compatible and an Anthropic-compatible endpoint.
The European tools below come from France, Germany, Latvia, Slovakia, Switzerland and Estonia. They range from open-weight models and per-token APIs to sovereign platforms that run inside your own infrastructure, under EU, Swiss or UK-style contracts rather than Chinese law.
How we rank these tools — 4-step process
-
1
European ownership, verified
The company is headquartered and incorporated in the EU, EEA or Switzerland, and processes customer data in Europe. A US parent company disqualifies a tool from this page regardless of where its servers are.
-
2
Category fit and hands-on review
What the tool actually does, who it suits, and where it falls short — checked against the vendor’s own documentation, changelog and pricing page rather than its marketing copy.
-
3
Compliance and pricing check
GDPR posture, hosting location and the prices quoted on this page are verified against the vendor’s public pricing before publication, and re-checked when we revisit the category.
-
4
Position on this page
Placement on this page can be paid, and that can affect which tools appear here and the order they appear in. It never buys a good review: a tool that fails the checks above is not here at any price, and payment does not change the shortcomings we write about. A vendor can ask us to correct a factual error — not to remove a criticism.
Vendors can pay for visibility on this page. It never changes what an entry says about a product, including the criticism, and we earn nothing when you click through to a vendor. Paid placement can affect which tools appear here and the order they appear in. Editorial policy
10 European Alternatives to DeepSeek
Mistral AI
Paris lab with open weights, reasoning models and a per-token API
Heabsy
Drop-in OpenAI and Anthropic API on EEA GPUs, zero retention
Lumo
Swiss-hosted chat app that runs open-weight models for you
Pleias
Small Apache 2.0 models trained on rights-cleared data
Aleph Alpha
Sovereign enterprise AI platform with source attribution
LightOn
Licence-priced private LLM platform for document work
Tilde
Open 30B multilingual model for Baltic and CEE languages
DentroChat
Browser assistant with a public list of European-only subprocessors
H Company
Holo models and agents that operate a browser or desktop
Kyutai
Open real-time voice models from a Paris non-profit lab
Key takeaways
- DeepSeek's Terms of Use, updated 27 March 2026, are governed by the laws of the People's Republic of China (mainland).
- DeepSeek's privacy policy states that it collects, processes and stores personal data in the People's Republic of China.
- On 20 October 2026 the DeepSeek API listed deepseek-flash from USD 0.15 per million uncached input tokens off-peak, and USD 0.30 at peak.
- Heabsy offers OpenAI- and Anthropic-compatible endpoints on dedicated GPUs in EEA data centres, so a DeepSeek client usually needs only a new base URL.
- Mistral AI, Pleias and Tilde publish models under open licences that you can host on your own servers in Europe.
Why people leave DeepSeek
DeepSeek is a Chinese AI lab whose models and apps have become a default for people who want strong reasoning at very low cost.
The consumer app and the developer API both run on its V4 generation, and the API documentation lists a Flash model with a one-million-token context window next to a Pro model, both with a thinking mode that is switched on by default. For a developer comparing price per million tokens, almost nothing else on the market looks as cheap, and that is the honest reason it is so widely tried.
The reasons to look beyond it are in its own legal documents. The Terms of Use say the services are owned and operated by Hangzhou DeepSeek Artificial Intelligence Co., Ltd., and section 10.1 puts the contract under the laws of the People's Republic of China (mainland).
The privacy policy names the same company as controller, with a registered address in China, and states that it collects, processes and stores personal data in the People's Republic of China. A European buyer who sends customer text through that service is therefore exporting it to a Chinese entity under Chinese contract law.
None of that says the models are bad. It says the contract, the court and the storage location sit outside the EU, EEA, UK and Switzerland, which is a hard stop for many procurement teams and for anyone handling personal data under GDPR.
The European tools below answer a narrower question than "is there a model as cheap as DeepSeek": where can you get comparable work done under a European contracting entity, and in some cases on weights you run yourself.
- Chinese law governs the contract Section 10.1 of the Terms of Use says disputes are governed by the laws of the People's Republic of China in the mainland, and section 10.2 lets either party sue in the court that has jurisdiction over the registered office of Hangzhou DeepSeek Artificial Intelligence Co., Ltd. For a European company, that means any dispute about data, outputs or billing is argued under foreign law at a foreign court. Mistral, Aleph Alpha, Tilde and the others below contract under the law of an EU member state or Switzerland.
- Personal data is stored in China The privacy policy, last updated 10 February 2026, states that DeepSeek directly collects, processes and stores personal data in the People's Republic of China, and mentions safeguards only "where required" when data moves out of certain countries. It lists prompts, uploaded files, chat history, device identifiers and approximate location among the data collected. Heabsy, Lumo and DentroChat publish where inference runs, and in each case it is inside Europe or Switzerland for the main service.
- The terms favour the provider The Terms of Use cap DeepSeek's liability at the greater of what you paid in the previous 12 months or one hundred dollars, provide the services "as is", and say shared chat links can be harvested by crawlers once published. They also reserve the right to change terms and prices. Those are common clauses, but a European buyer negotiating a data processing agreement with LightOn or Aleph Alpha gets a negotiated contract, which a standard click-through cannot offer.
- Peak-hour pricing and a moving model line-up The API pricing page lists two rates for each token class, with off-peak at half the peak rate, and defines peak as 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays. It also notes that legacy model names are retired and silently served by the newer Flash model. Cheap is real, but cost depends on when you call it, and a model you validated can be replaced under the same name, which matters for regulated workflows.
What you have to replace, not just match
What you are replacing is a price point and a habit. Teams using DeepSeek's API typically point an OpenAI SDK or a coding agent at the base URL https://api.deepseek.com, or at the Anthropic-format URL under /anthropic. Both European APIs that offer that same compatibility, Heabsy most directly, make it a base-URL change. For the other European APIs, plan to check client code and prompts against each vendor's own documentation, which we did not review here.
If you use the DeepSeek chat app, you are replacing a free assistant with a deep-thinking mode. Lumo and DentroChat are the nearest European apps, and Le Chat from Mistral is the closest in capability. If you run open weights yourself, the replacement is a different set of weights, not a different vendor: Mistral's Apache 2.0 models, Pleias's small models and Tilde's TildeOpen-30B can all be downloaded and hosted on your own servers, which removes the data-location question entirely.
Finally there is the part people forget, the surrounding services. DeepSeek also sells a Search and Fetch API at USD 0.75 per 1,000 successful requests, and its pricing page notes that server-side search and fetch calls made through its Harness, Claude Code or Codex integrations are free until billing starts on 1 November 2026. None of the European tools sells an exact equivalent of those, so check whether your agent setup depends on them before you move.
The alternatives compared
| Position | Tool | Headquarters | Pricing | Jurisdiction |
|---|---|---|---|---|
| #1 | Mistral AI | Paris, France | Free tier / pay-per-use API | EU (France) |
| #2 | Heabsy | Bratislava, Slovakia | Per million tokens in US dollars, for example gpt-oss-20b at $0.05 input and $0.20 output / no minimum spend or subscription | EU (Slovakia) |
| #3 | Lumo | Geneva, Switzerland | Free tier with limits / Lumo Plus $9.99 per month billed annually / Lumo for Business via Lumo Professional or Proton Workspace | Switzerland |
| #4 | Pleias | Paris, France | Common Corpus dataset and Pleias small language models (Pleias-RAG-1B, Pleias-Pico and others) free and open on Hugging Face under Apache 2.0; Synth and Stratum on-premise deployments priced through a demo call, no published rate card. | EU (France) |
| #5 | Aleph Alpha | Heidelberg, Germany | Enterprise pricing on request | EU (Germany) |
| #6 | LightOn | Paris, France | Enterprise licence pricing on request | EU (France) |
| #7 | Tilde | Riga, Latvia | Machine translation, speech, AI assistant and TildeOpen LLM products priced per enterprise engagement, quoted on request; TildeOpen-30B released free and open under CC BY 4.0 on Hugging Face. | EU (Latvia) |
| #8 | DentroChat | Tallinn, Estonia | Free tier of 15 messages / EUR 1 seven-day trial / EUR 12 per month or EUR 97 per year for full access | EU (Estonia) |
| #9 | H Company | Paris, France | HoloTab Chrome extension free; Holo Models API self-serve via hub.hcompany.ai; Autonomous Enterprise Operations platform and custom model access priced through direct sales engagement. | EU (France) |
| #10 | Kyutai | Paris, France | Free and open source models | EU (France), non-profit |
How each alternative compares to DeepSeek
Mistral AI
The nearest European counterpart for reasoning models, open weights and a developer API
Best for: Developers who want a reasoning model, open weights and an API in one European vendor.
Against DeepSeek, Mistral AI is the vendor that covers the same ground in one place. Its Magistral Small and Medium models are the reasoning line, with the Small version open-sourced, and Mistral Large 3 is a mixture-of-experts model with 41 billion active of 675 billion total parameters and a 256,000-token context, released as an open-weight download. That context is a quarter of what DeepSeek lists for its Flash model, which is the clearest capability gap on paper.
Mistral AI is a Paris company, so API calls fall under French law and EU data protection rules instead of Chinese mainland law. Le Chat is the consumer assistant, free with a Pro plan at USD 14.99 a month, and La Plateforme is the API, with prices from roughly EUR 1 per million input tokens for Mistral Small. Self-hosting an open model costs only compute.
What Mistral AI does better than DeepSeek
- Contract and court under French and EU law rather than Chinese mainland law
- Open-weight models you can host yourself, including a frontier-class one
- Le Chat gives non-developers an app with deep research and memory
- Full range from 3B edge models to a 675B-parameter model
Where Mistral AI is a step down from DeepSeek
- Published per-token prices are higher than DeepSeek's Flash rates
- 256,000-token context against the one million DeepSeek lists for Flash
- Not every model in the line-up is open-weight
Standout against DeepSeek. Frontier-class weights you can download and run inside your own European infrastructure.
Best for: Teams running coding agents or SDK code who want to change only the base URL.
Heabsy is the closest in shape to how most developers use DeepSeek: an HTTP API with OpenAI- and Anthropic-compatible endpoints, billed per token with no minimum spend. The company, FEYA, s.r.o., is registered in Bratislava with no US parent. Its flagship model, Qwen3.8 27B, runs on dedicated GPUs in EEA data centres, and the data processing agreement states that prompts and completions are not logged and nothing is used for training.
Its catalog also lists models from other providers, DeepSeek included, fulfilled through global providers under Heabsy's European contract. That is a way to keep a DeepSeek model while contracting with a Slovak company, although those routed models may be processed outside the EEA. The flagship costs $0.40 per million input tokens and $3.00 per million output tokens, higher than DeepSeek's Flash rates.
What Heabsy does better than DeepSeek
- Zero retention and EEA-only compute for the flagship model, in a signed DPA
- One European invoice that can also cover routed models, DeepSeek among them
- Per-key budgets and early rate-limit signals instead of a silent queue
- Published throughput and cache-hit figures rather than price alone
Where Heabsy is a step down from DeepSeek
- Output tokens at $3.00 per million cost far more than DeepSeek Flash output
- Only one model carries the EEA-only guarantee; routed models can leave the EEA
- Young platform with no published founding date and no free tier
Standout against DeepSeek. A base-URL change gets an existing DeepSeek client onto EEA hardware with zero retention.
Best for: Individuals and Proton users who want a private assistant, not a developer endpoint.
Lumo is Proton's assistant, and it replaces the chat side of DeepSeek rather than the API. Chats use zero-access encryption, which means Proton cannot read your history, and the guest and Tor access options let you use it without an account. Proton AG is in Geneva, so the contract is Swiss, and inference runs on Proton-operated servers in Switzerland.
One point matters for DeepSeek users in particular. Lumo's tool page says it runs open-weight models, currently Qwen 3.5 and GLM 5.2 for text, which were trained by Chinese labs. The weights are open and inference happens in Switzerland, so your prompts do not go to China, but the models themselves are not European-made. Pricing is a free tier, Lumo Plus at $9.99 a month billed annually, or Lumo for Business.
What Lumo does better than DeepSeek
- Zero-access encrypted history that Proton cannot read
- Guest and Tor access without an account
- Apps and models are open source and inspectable
- Reads files from Proton Drive if you already use the suite
Where Lumo is a step down from DeepSeek
- No API or self-hosting, so it cannot replace DeepSeek in a pipeline
- Behind frontier models on hard reasoning and coding
- The base models were trained by Chinese labs, which some buyers rule out
Standout against DeepSeek. Prompts are processed in Switzerland even though the underlying weights were trained in China.
Best for: Teams that need a small, auditable model for retrieval or document tasks on their own hardware.
Pleias answers DeepSeek's low-cost appeal differently: instead of a cheap API, it publishes small language models that cost nothing per token because you run them. Pleias-RAG-1B and the Pico family are on Hugging Face under Apache 2.0, and the site describes them as small enough to run offline on hardware costing under EUR 100. They were trained on Common Corpus, a dataset of rights-cleared material with documented provenance.
Pleias SAS is a Paris company founded in 2023. Its Synth and Stratum products for synthetic data and document processing deploy on-premise, and are priced through a demo call. This is not a general chat or reasoning replacement. Expect a model that does one defined job, such as retrieval-augmented answers over your own documents, and not open-ended multi-step problem solving.
What Pleias does better than DeepSeek
- Models run offline on your hardware, so no prompt leaves your network
- Training data with documented, rights-cleared provenance
- Apache 2.0 licence on the small models
- RATP's Sillon model shows a production use in transport
Where Pleias is a step down from DeepSeek
- No general-purpose flagship or chat product
- No reasoning mode or large context comparable to DeepSeek V4
- Synth and Stratum have no public price list
Standout against DeepSeek. Small models and a rights-cleared dataset that you can inspect and run entirely on your own machine.
Aleph Alpha
A sovereign platform for regulated organisations, priced by contract rather than by token
Best for: Public bodies and regulated enterprises that need AI inside a controlled boundary.
Aleph Alpha sits at the other end from DeepSeek's cheap public API. PhariaAI is an enterprise platform covering model hosting, orchestration, retrieval and compliance, running on STACKIT so that compute and software are both operated by European companies. Its hybrid execution splits workloads by data classification, keeping sensitive material on local servers and letting less sensitive tasks use cloud capacity.
Aleph Alpha GmbH is in Heidelberg and sells on an enterprise basis, with pricing on request and proprietary models instead of open weights. Source attribution shows which documents contributed to an output, which helps with EU AI Act review. Expect a sales process, not a self-service key, and expect raw capability below the largest models.
What Aleph Alpha does better than DeepSeek
- Source attribution that makes outputs traceable for audit
- On-premise and hybrid deployment for classified workloads
- German contracting entity and a European operating stack
Where Aleph Alpha is a step down from DeepSeek
- No published pricing and no self-service entry point
- Proprietary models with no open weights, unlike DeepSeek's downloadable releases
- Raw capability sits below the largest foundation models
Standout against DeepSeek. Hybrid execution decides per data class what stays on-premise and what may use the cloud.
Best for: French and European enterprises running agents over large internal document sets.
LightOn's Paradigm runs on-premise or in a European private cloud and serves LightOn's own models next to open-source ones through one interface. For a DeepSeek user, the shift is from metered tokens to a licence fee: costs are a figure you can budget, not one that moves with peak and off-peak hours. Fine-tuning happens inside your own infrastructure, so training data and the tuned model stay with you.
LightOn SAS is a Paris company that spun out of École Polytechnique research. The platform's agents search document bases, extract information and compare long sets of documents. It is sold to enterprises, so a developer who wants a quick key to test a prompt will not find a self-service route, and the largest foundation models remain American or Chinese.
What LightOn does better than DeepSeek
- Predictable licence cost against per-token billing
- Fine-tuned models never leave your infrastructure
- French contracting entity with reference clients in the public sector
Where LightOn is a step down from DeepSeek
- Enterprise-only, with pricing on request
- Deployment is heavier than calling a hosted API
Standout against DeepSeek. The bill is a licence you can plan, with nothing metered by the token.
Best for: Public-sector and regulated teams working in Latvian, Lithuanian, Estonian and similar languages.
Tilde is a different kind of alternative: a language-technology firm in Riga, active since 1991, whose open model TildeOpen-30B is released under CC BY 4.0 on Hugging Face. For a team that picked DeepSeek for general use and now needs strong handling of smaller EU languages, TildeOpen is a model you can download, inspect and host in Europe, which no hosted Chinese API allows.
The company, SIA "Tilde", also sells machine translation, speech and assistants on enterprise terms, and holds ISO 27001 certification. Pricing for those is on request. TildeOpen-30B is not built to beat general models on open-ended reasoning, so judge it on the languages it covers and not against DeepSeek's headline tasks.
What Tilde does better than DeepSeek
- Coverage of Baltic and Central European languages the larger labs undertrain
- Open download under CC BY 4.0 for self-hosting in Europe
- ISO 27001 certification for public-sector procurement
Where Tilde is a step down from DeepSeek
- Weaker general reasoning than DeepSeek or Mistral
- Managed products have no public price list
Standout against DeepSeek. Three decades of Baltic and CEE language work behind a model you can download.
Best for: European organisations that must show a procurement team exactly which companies touch the prompts.
DentroChat replaces the DeepSeek chat app with a web assistant that has three answer modes, Fast, Thinking and Creative, along with image generation, web search, file analysis, voice input and a document workspace. The point against DeepSeek is traceability: its published list of eighteen subprocessors across nine European countries contains no American or Chinese name.
DentroInnovation OÜ in Tallinn does not train its own models. It assembles European providers behind one interface, so quality depends on them and can change when they do. Pricing is a free tier of 15 messages, a EUR 1 seven-day trial, then EUR 12 a month or EUR 97 a year. There is no self-hosting and no public API on the tool page.
What DentroChat does better than DeepSeek
- A public subprocessor list you can hand to a procurement team
- Document, deck and spreadsheet output from a conversation
- Automatic chat deletion settings for privacy-minded users
Where DentroChat is a step down from DeepSeek
- No in-house models, so quality follows whatever its providers offer
- Small company with far less scale than Mistral or Proton
- Web only, with no self-hosting, open source or developer endpoint
Standout against DeepSeek. Every company processing your prompts is named, and all of them are European.
Best for: Teams automating web and desktop tasks who are testing agent models beyond plain text.
H Company is relevant if your DeepSeek use is agentic rather than conversational. Its Holo models interpret screens and carry out multi-step tasks, HoloTab is a free Chrome extension with recurring Routines, and the Holo Models API is available through a self-serve hub. Computer-use agents were added in July 2026 to act on a whole computer.
The Paris company was founded in 2024 and sells its Autonomous Enterprise Operations platform through direct sales, with no public price list beyond the free extension. Its models are proprietary rather than open-weight, and it competes with agent products from much larger labs, so test it on your own workflows before you commit.
What H Company does better than DeepSeek
- Free browser agent to try without a contract
- French contracting entity for agent workloads
Where H Company is a step down from DeepSeek
- No open weights and no public pricing for the API or enterprise platform
- Short track record since its 2024 founding
Standout against DeepSeek. It targets software that acts on a screen, not a chat window that describes the steps.
Best for: Engineers building voice interfaces who want weights they can run and inspect.
Kyutai is on this page for voice, not for text. DeepSeek's API pricing page lists text models, with vision on the Flash model, and no speech model. Kyutai's Moshi speaks and listens at once, with a practical latency of 200 to 240 milliseconds, and the lab releases the models openly so they can run locally with no per-minute fee.
Kyutai is a Paris non-profit research lab funded with EUR 300 million from French backers, not a vendor. There is no interface, no support and no service-level agreement, and using the models takes engineering effort. It will not replace DeepSeek for chat, coding or reasoning, and it is last in the ranking for that reason.
What Kyutai does better than DeepSeek
- Real-time conversational voice that runs on your own machine
- Open release with no metered API
Where Kyutai is a step down from DeepSeek
- Voice-focused, not a general text or reasoning model
- No product, support or service-level agreement
Standout against DeepSeek. A downloadable voice model with latency low enough to feel like conversation.
What actually breaks when you switch
The first thing that breaks is the price model. DeepSeek bills uncached input, cached input and output separately, with a peak and an off-peak rate, and its Flash model lists a 1M-token context and up to 384K output. Moving to a European API changes all four numbers at once, so rerun your own token counts through each vendor's published rates before you decide, and check whether your prompts rely on long contexts that a 256,000-token window will not hold.
The second is behaviour. Thinking mode is on by default at DeepSeek, so prompts tuned for it may need rewriting for a model that reasons differently, and output formats for tool calls vary between vendors.
Exporting is easier than usual, because DeepSeek's privacy policy grants a right to a portable copy of your data, but chat history is not a standard format. Plan a test set of real prompts, run it on two candidates, and keep DeepSeek running in parallel until the answers match.
Can I keep using DeepSeek models but contract with a European company?
Partly. The Heabsy tool page says its catalog includes models from several providers, DeepSeek among them, fulfilled through global providers under Heabsy's own European contract. That gives you one invoice and one data processing agreement with a Slovak company. The same page is explicit that these routed models may be processed outside the EEA under the provider's own policy.
Only Heabsy's flagship model, Qwen3.8 27B, carries the EEA-only and zero-retention guarantee. So the answer depends on what you need: a European contract for a DeepSeek model, or European processing. You can get the first for DeepSeek. For the second, choose the flagship model or a model you host yourself.
Is any European model as strong at reasoning as DeepSeek?
Mistral's Magistral line is its reasoning family, with Magistral Small released as open weights and Mistral Large 3 offered as a frontier-class open-weight model. On the benchmarks you care about, the only honest way to know is to run your own tasks on both. We have not measured the models against each other, and we do not publish a ranking.
What is clear is the cost gap. DeepSeek's published per-token rates are low, and the European APIs listed here are not trying to undercut them. If your workload is high-volume classification or bulk summarisation where reasoning depth matters less, the price difference will dominate, and a small open model such as Pleias-RAG-1B on your own hardware can be the cheaper European route.
Which one to pick
For most readers, start with Mistral AI: it is the only European vendor here with a reasoning model family, open weights and a public API, and its Paris base gives you EU law.
If your code already speaks the OpenAI or Anthropic protocol and you want the least work, try Heabsy, and decide whether you need its flagship model for the EEA-only guarantee or a routed model for variety. For a private chat app, pick Lumo if you value encryption and Proton integration, or DentroChat if you need the subprocessor list.
If the driver is sovereignty, go to the self-hosted end: Pleias for small auditable models, Tilde for Baltic and CEE languages, Aleph Alpha or LightOn when a regulated organisation wants a supported platform and accepts enterprise pricing. Choose H Company only for agent automation and Kyutai only for voice. If you only need cheap bulk tokens and your data is not personal, DeepSeek may still be the cheapest option, and you should say so in your own comparison.
Frequently Asked Questions
For most developers and businesses, Mistral AI is the closest match: a Paris lab with a reasoning model family, open-weight downloads and a per-token API under French law. Heabsy is the easiest swap for code that already calls an OpenAI- or Anthropic-style endpoint, since it runs open models on dedicated GPUs in EEA data centres. For a chat app rather than an API, Lumo from Switzerland and DentroChat from Estonia are the options on this page.
The main reason is where the contract and the data sit. DeepSeek's Terms of Use are governed by the laws of the People's Republic of China (mainland), and its privacy policy says personal data is stored in China. Companies that process personal data of EU residents, or that must keep data under EU or Swiss law, often cannot accept that. Price and model quality are rarely the objection.
You can run open-weight models of your own choice. Mistral publishes Apache 2.0 models, Pleias publishes small models under Apache 2.0 on Hugging Face, and Tilde released TildeOpen-30B under CC BY 4.0. Hosting them yourself keeps prompts on your own infrastructure. Whether a particular DeepSeek release may be self-hosted is a licence question we did not verify, so read its licence before you plan around it.
Usually, per token. DeepSeek's pricing page lists deepseek-flash at USD 0.15 per million uncached input tokens off-peak, while Heabsy lists USD 0.40 per million input tokens for its flagship model, and Mistral lists roughly EUR 1 per million input tokens for Mistral Small according to the site's tool review. Licence-priced platforms such as LightOn trade per-token cost for a predictable fee.
Not automatically. A European contracting entity gives you European contract law, but compute can still run elsewhere, as Heabsy's own page notes for routed catalog models. Ask each vendor where inference runs, whether prompts are logged and whether a data processing agreement is available. Self-hosting open weights is the only option on this page that removes the question completely.
Related comparisons
About the author
Explore More European Alternatives
Discover privacy-focused European alternatives to other popular US tech services.