Хиймэл оюун ухааны салбарын томоохон компаниуд их хэмжээний хөрөнгө оруулалт хийж буй энэ үед LinkedIn өөрийн дата төвийн хүчин чадлыг нэмэгдүүлэхгүй байх стратеги баримталж байна.
Тус компани ирэх зургаадугаар сард дуусах санхүүгийн жилд GPU болон өгөгдөл хадгалах дэд бүтцийн зардлаа өсгөхгүй байхаар төлөвлөжээ. Сүүлийн зургаан сарын хугацаанд одоо байгаа GPU-ийн ашиглалтын үр ашгийг хоёр дахин нэмэгдүүлснээр техник хангамжийн нэмэлт зардал гаргах шаардлагагүй болсон гэж удирдлагууд мэдээлэв. Инженерийн багийн ахлах технологийн захирал Эрран Бергерийн хэлснээр, компани нь илүү их тооцоолол шаардсан бүтээгдэхүүнүүдийг зах зээлд гаргахын зэрэгцээ дата төвийн хүчин чадлыг тогтвортой байлгах зорилт тавьжээ.
LinkedIn-ийн дэд бүтцийн асуудал хариуцсан технологийн захирал Рагу Хиремагалур энэхүү хэмнэлт нь инженерийн багуудад хиймэл оюун ухааны шинэ функцүүдийг хөгжүүлэхдээ илүү бүтээлч байх сэдэл өгнө гэж үзэж байна. Тус компани 2022 оноос хойш Орегон, Техас, Виржиниа мужид өөрийн дата төвүүдийг бие даан ажиллуулж эхэлсэн нь технологийн хяналтаа сайжруулах боломжийг олгосон юм. Энэ нь хиймэл оюун ухааны туслах хэрэгслүүдийн ашиглалтын өртөг өсөж буй өнөөгийн нөхцөлд зардлаа оновчтой удирдах чухал алхам болж байна.
OpenAI, Meta, Google зэрэг компаниуд дата төвийн хүчин чадлаа өргөжүүлэхийн тулд асар их хөрөнгө зарцуулж байгаа энэ үед LinkedIn-ийн энэхүү шийдвэр салбарынхны анхаарлыг татаж байна. Principal Venture Partners-ийн удирдах түнш, HP компанийн удирдах зөвлөлийн гишүүн Сонги Юн үүнийг хиймэл оюун ухааны технологи туршилтын шатнаас үйлдвэрлэлийн түвшинд шилжиж буйн дохио хэмээн дүгнэв. Түүний үзэж буйгаар, ирээдүйд зөвхөн дэд бүтцэд хамгийн их хөрөнгө оруулсан нь бус, харин үр ашигтай ажилласан компаниуд зах зээлд ялагч болох юм.
Дэлгэрэнгүйг эх сурвалжаас харах
Эх сурвалжийг нээх ↓
Unlike many other big tech platforms, LinkedIn has decided it won’t spend aggressively on expanding its AI data centers this fiscal year. Executives at the professional social network tell WIRED that it plans to keep its investment in GPUs steady, and its compute and storage footprint is also remaining flat.
The spending calculations apply to LinkedIn’s fiscal year that began last month and ends next June. The company says it was able to avoid spending big on AI hardware because it found ways to use its existing GPUs twice as efficiently over the past six months. LinkedIn’s plan could still unravel because the hardware demands of AI are shifting rapidly, but executives say the company has already taken into account surging prices for memory chips.
“One of the goals we’ve set is to try to basically keep our compute footprint flat or as close to flat as possible while shipping more compute-hungry things to production,” says Erran Berger, LinkedIn’s chief technology officer for engineering. “That’s a pretty bold statement to make in today’s world.”
Berger and Raghu Hiremagalur, LinkedIn’s chief technology officer for infrastructure, say they want to be prudent about spending and that the new constraints will motivate engineering teams to get more creative when developing the many new generative AI features LinkedIn is planning to launch. Berger says he believes the efficiency gains could compound over time, enabling LinkedIn to get more out of data center expansions when it eventually increases its budgets again.
“I really want to double underscore that for a company of our scale, to say a full year we’re going to do this with no incremental storage and compute is no small feat, but it’s taken a ton of work to get there,” Hiremagalur says.
Companies such as OpenAI, Meta, and Google are scrounging up all the money they can find and coupling up in unexpected partnerships to construct, furnish, and operate massive data centers filled with the newest computer chips. Labor and parts shortages have held up many projects, and many businesses have had to limit customer usage of some AI tools. But there are also growing questions about whether the relentless investment in AI is sustainable. LinkedIn, with more than 1.3 billion users, is perhaps the largest business yet to publicly address spending concerns by bucking the building boom.
“It is encouraging for the industry,” says Songyee Yoon, managing partner of Principal Venture Partners and a board member at the server maker HP. “It suggests AI is beginning to move from experimentation into production discipline. The companies that win will not simply be the ones that spend the most on infrastructure.”
Owning It
A few years after Microsoft acquired LinkedIn in 2016, the company tried moving to its parent company’s Azure cloud service, but it didn’t make economic sense to squeeze the giant social network into general-purpose data centers. “Microsoft Azure was growing like crazy, the level of customer demand was through the roof, and at the same time we saw skyrocketing growth on the LinkedIn side,” Hiremagalur says.
In 2022, LinkedIn went all-in on its own data centers in Oregon, Texas, and Virginia. The ownership gave LinkedIn significant control over every detail of its technology, setting itself up well to meet the realities of a new era. Around the same time, LinkedIn began developing AI-based assistants that could help users write messages, find jobs, and recruit candidates. The endeavor wasn’t cheap. “Every query that’s coming to our site has increased in cost over time,” Hiremagalur says, adding that the amount of data LinkedIn stored was doubling annually. “That is not a sustainable place to be.”
