一家位于前袜子工厂旧址的波罗的海科技公司，如今已迅速崛起，成为立陶宛最新的“独角兽”企业，并确立了自身作为数据驱动人工智能革命中核心引擎的地位。
A Baltic tech company tucked away on the site of a former sock factory has soared to become Lithuania’s latest unicorn, positioning itself as the workhorse of the data-driven artificial intelligence revolution.
总部位于维尔纽斯的Source是一家网络数据集团，于周二正式启用，成为Oxylabs及另外四家网络情报公司的母公司。首席执行官Vytautas Savickas表示，随着人工智能开发者对公开在线数据的需求激增，公司正在快速成长。为满足需求，公司已将数据处理能力提升了十倍。
Vilnius-based Source, a web-data group that launched Tuesday as the umbrella group for Oxylabs and four other web-intelligence companies, is growing rapidly as AI developers fuel demand for public online data, Chief Executive Officer Vytautas Savickas said. The company has increased its data-processing capacity tenfold to keep pace with demand.
“这是前所未有的最大浪潮，”Savickas在接受彭博新闻采访时说道。“基本上，十大大型语言模型实验室中有八家都依赖我们的基础设施来实时收集数据。”他指的是大型语言模型（LLM），但出于保密协议，他拒绝透露具体名称。
“It’s the biggest wave ever,” Savickas said in a Bloomberg News interview. “Basically, eight out of 10 biggest LLM labs leverage our infrastructure to collect that data real time,” he said, referring to large language models. He declined to name them, citing non-disclosure agreements.
今年夏天，在获得Warburg Pincus的注资并实现36亿美元估值后，该集团成为立陶宛第六家“独角兽”企业——即估值达到或超过10亿美元的私营公司。这也使其成为仅次于在线时尚平台Vinted的立陶宛第二高价值的私营科技公司。
This summer, the group became Lithuania’s sixth unicorn — a privately held company worth $1 billion or more — after securing investment from Warburg Pincus at a valuation of $3.6 billion. That makes it the country’s second-most valuable private technology company behind online fashion marketplace Vinted.
Source由2015年成立的网络爬虫公司Oxylabs发展而来。Oxylabs已帮助《财富》500强企业在电子商务、旅游和网络安全领域大规模收集和处理公开在线信息。
Source grew out of Oxylabs, a web-scraping company founded in 2015, which has helped Fortune 500 companies in e-commerce, travel and cybersecurity collect and process online public information at scale.
Savickas表示，全球约有35万个技术团队在使用Source的解决方案，这些团队规模从五人初创公司到千人级巨头不等，公司年经常性收入已达3.5亿美元，其中一半来自美国市场。为更贴近关键客户，包括科技公司、顶级零售商和人工智能实验室，公司将在旧金山开设办事处。
Some 350,000 tech teams, ranging from five-person startups to 10,000-person powerhouses, use Source solutions globally, generating $350 million in annual recurring revenue run rate, half of which come from the US market, Savickas said. The company will open a San Francisco office to be closer to key clients, including tech companies, top retailers and AI labs.
为了体现其发展的新阶段，该集团采用了Source这一名称，同时保留Oxylabs作为独立品牌。
In a nod to its next phase, the group adopted the Source name while retaining Oxylabs as a separate brand.
这家公司正在开发一套全新的网络应用程序接口，该接口将依托其自主构建的互联网索引系统，把现有的各类数据收集工具整合为一项专为人工智能代理设计的统一服务。
The firm is in the process of building a new web API, or application programming interface, which will be backed by its own index of the internet to consolidate all of its existing data-collection tools into a single service designed specifically for AI agents.
该服务旨在帮助人工智能代理快速从网络上检索最新的公开信息。目前产品仍处于早期测试阶段，预计将在未来几周内正式发布。整个开发工作由一支约100名工程师和开发人员组成的团队独立完成。
The product is intended to let AI agents rapidly search and retrieve fresh public information from the web. Currently in early beta, it is set to launch in the coming weeks and has been built entirely in-house by a team of about 100 engineers and developers.
萨维卡斯在谈到人工智能模型时指出：“它们需要最新鲜的数据作为回答的依据，否则就可能出现‘幻觉’现象，即生成错误或不合理的信息。”
“They need the freshest data to ground their answers, otherwise they can hallucinate,” Savickas said of the AI models.
萨维卡斯是在维尔纽斯“网络城”园区内的Source总部接受采访的。该园区是立陶宛众多科技企业的聚集地：Nord Security公司就在隔壁，Vinted公司则位于街对面。园区附近还有一个名为“Vienaragių”的公交站，在立陶宛语中意为“独角兽”。
Savickas was speaking from Source’s headquarters located inside Vilnius’s Cyber City campus, a dense cluster that brings together Lithuania’s biggest tech names. Nord Security is next door and Vinted is across the street. The area is served by a bus stop named Vienaragių — Lithuanian for unicorn.
这些大楼原先是一座袜子工厂的所在地。如今，这一街区已成为立陶宛从上世纪90年代的低成本制造业经济体转型为新兴科技中心的象征，数字服务在该国出口中的比重也日益提升。
The buildings were constructed on the site of a former sock factory. The single city-block has come to symbolize Lithuania’s transformation from a low-cost manufacturing economy in the 1990s into a growing technology hub, with digital services playing an increasingly important role in the country’s exports.
萨维卡斯透露，今年上半年，Source的数据处理能力提升了约十倍，如今每天可处理“数以拍字节计”的数据量，以应对人工智能领域需求的激增。处理如此庞大的数据量需要持续进行资源管理。
Source increased its data-processing by about tenfold in the first half of this year, enabling it to process “petabytes and petabytes” of data a day as AI demand surges, Savickas said. Handling the mountain of data requires constant resource management.
今年刚出任首席执行官的萨维卡斯坦言：“你总会感觉自己落后了，这让人倍感焦虑。”为克服技术难题，公司招聘了数百名员工，目前全球员工总数已达800人，其中开发人员与工程师超过350名。Source的主要竞争对手是以色列的Bright Data，这两家公司均是全球最大的网络数据基础设施供应商。
“You always feel behind,” said Savickas, who took over as CEO this year. “It brings a lot of anxiety.”The company hired hundreds of employees to overcome engineering challenges, and now employs 800 globally, including more than 350 developers and engineers. Source competes most directly with Israel’s Bright Data, with the two companies among the largest global providers of web-data infrastructure.
除了能否满足数据需求之外，网络上信息的可获取性对 Source 公司的业务发展也至关重要。随着人工智能驱动的数据抓取技术（AI-driven scraping）的兴起，相关企业遭到了批评和诉讼；这些诉讼指控这些公司未经许可就提取了用户数据，从而损害了内容创作者的权益，并引发了版权方面的争议。
Beyond the ability to keep up with data demand, the future accessibility of information on the web is also key for Source’s business. The rise of AI-driven scraping has triggered criticism and lawsuits against AI companies, alleging the extraction of information without compensation, thereby undermining content businesses and raising copyright concerns.
Savickas 表示，Source 公司始终致力于维护开放的网络环境，同时也强调企业有权保护用户的私人数据及知识产权。Source 公司制定了相应的合规与道德规范，以确保其数据抓取行为不会侵犯这些权益。
Source is a defender of the open web, Savickas said, adding that companies also have a right to protect private data and intellectual property. Source has compliance and ethics rules to ensure its web-scraping respects those boundaries, the Savickas said.
首席执行官认为，限制网络信息的流动实际上会损害那些依赖公共数据来开发产品、服务及人工智能工具的中小企业的利益，反而会让大型科技公司受益。
Limiting the flow of information online would benefit the biggest technology companies at the expense of businesses that rely on public data to build products, services and AI tools, according to the CEO.
“大型科技公司总是试图垄断数据，”Savickas 说，“这虽然会让这些巨头受益，但并不会促进整个经济的发展。”
“Big tech, big players want to monopolize” data, Savickas said. That would be “benefiting the core big tech, but then not the whole economy.”