• Topics
  • Artificial wisdom
  • Autopilot
  • network
  • Processor
  • 手機
  • exhibition activities
    • CES
      • CES 2014
      • CES 2015
      • CES 2016
      • CES 2017
      • CES 2018
      • CES 2019
      • CES 2020
    • MWC
      • MWC 2014
      • MWC 2015
      • MWC 2016
      • MWC 2017
      • MWC 2018
      • MWC 2019
    • Computex
      • Computex 2014
      • Computex 2015
      • Computex 2016
      • Computex 2017
      • Computex 2018
      • Computex 2019
    • E3
      • E3 2014
      • E3 2015
      • E3 2016
      • E3 2017
    • IFA
      • IFA 2014
      • IFA 2015
      • IFA 2016
      • IFA 2017
    • TGS
      • TGS 2016
  • About us
    • About mashdigi
    • mashdigi website contact details
Thursday, July 2026, 05, 21:18 AM
  • Login
mashdigi-Technology, new products, interesting news, trends
  • Topics
  • Artificial wisdom
  • Autopilot
  • network
  • Processor
  • 手機
  • exhibition activities
    • CES
      • CES 2014
      • CES 2015
      • CES 2016
      • CES 2017
      • CES 2018
      • CES 2019
      • CES 2020
    • MWC
      • MWC 2014
      • MWC 2015
      • MWC 2016
      • MWC 2017
      • MWC 2018
      • MWC 2019
    • Computex
      • Computex 2014
      • Computex 2015
      • Computex 2016
      • Computex 2017
      • Computex 2018
      • Computex 2019
    • E3
      • E3 2014
      • E3 2015
      • E3 2016
      • E3 2017
    • IFA
      • IFA 2014
      • IFA 2015
      • IFA 2016
      • IFA 2017
    • TGS
      • TGS 2016
  • About us
    • About mashdigi
    • mashdigi website contact details
No Result
View All Result
  • Topics
  • Artificial wisdom
  • Autopilot
  • network
  • Processor
  • 手機
  • exhibition activities
    • CES
      • CES 2014
      • CES 2015
      • CES 2016
      • CES 2017
      • CES 2018
      • CES 2019
      • CES 2020
    • MWC
      • MWC 2014
      • MWC 2015
      • MWC 2016
      • MWC 2017
      • MWC 2018
      • MWC 2019
    • Computex
      • Computex 2014
      • Computex 2015
      • Computex 2016
      • Computex 2017
      • Computex 2018
      • Computex 2019
    • E3
      • E3 2014
      • E3 2015
      • E3 2016
      • E3 2017
    • IFA
      • IFA 2014
      • IFA 2015
      • IFA 2016
      • IFA 2017
    • TGS
      • TGS 2016
  • About us
    • About mashdigi
    • mashdigi website contact details
No Result
View All Result
mashdigi-Technology, new products, interesting news, trends
No Result
View All Result
Home Life

ByteDance is reportedly set to spend a staggering $140 billion next year to acquire NVIDIA chips! However, they haven't given up on developing their own, with performance rumored to match H20 performance.

Author: Mash Yang
2025-12-31
in Life
A A
0
Share to FacebookShare on TwitterShare to LINE

The AI ​​computing power race is burning through cash! Despite the shadow of US export controls, TikTok's parent company, ByteDance, seems to have prepared its silver bullets to massively expand its capabilities in 2026.

ByteDance is reportedly set to spend a staggering $140 billion next year to acquire NVIDIA chips! However, they haven't given up on developing their own, with performance rumored to match H20 performance.

This is an advertisement.

According to the South China Morning PostGet the messageByteDance plans to allocate approximately 1000 billion RMB (about $140 billion) in 2026 specifically for purchasing AI chips from NVIDIA. This figure represents a significant increase from 850 billion RMB in 2025, demonstrating that its thirst for computing power shows no signs of abating.

As long as the H200 can be sold, money is not a problem.

Sources familiar with the matter revealed that this $140 billion budget is based on the premise that "NVIDIA is allowed to sell H200 GPUs to China." The H200 is one of NVIDIA's most powerful AI chips to date.US government approves (Or provide a downgraded version), and ByteDance will place an order without hesitation.

This is just one part of ByteDance's massive AI spending plan for 2026. It is rumored that its overall AI capital expenditure plan will reach 1600 billion RMB, including not only chip purchases but also infrastructure construction and talent recruitment.

What if you can't buy it? Reports indicate that their self-developed chip has successfully completed tape-out.

Of course, you can't put all your eggs in one basket. While spending money to acquire NVIDIA, ByteDance also has an internal chip design team of about 1000 people, quietly working on "Plan B".

The report indicates that ByteDance's internal chip division has achieved a major breakthrough, successfully tape-out a self-developed processor. This chip is said to have performance comparable to NVIDIA's H20 chip, which was downgraded for the Chinese market, but at a lower cost. This means that ByteDance has achieved a certain degree of self-sufficiency in entry-level or inference computing power.

In addition, the company is also actively developing HBM (High Bandwidth Memory) technology through internal R&D and investment in startups, attempting to master key components of AI chips.

The demand for steamed buns has surged, and Volcano Engine will be featured on the Spring Festival Gala.

Why does ByteDance need so much computing power? Just look at its AI chatbot "Doubao" and you'll know.

This is an advertisement.

Data shows that Doubao's daily token processing volume has exceeded 50 trillion this month, a remarkable increase compared to 4 trillion in December last year. Furthermore, its cloud service, Volcano Engine, has secured an exclusive AI cloud partner contract for CCTV's Spring Festival Gala, which will face extremely high traffic demands.

Geopolitical risk aversion: Is the chip industry "migrating" to Singapore?

It's worth noting that ByteDance appears to be adjusting its organizational structure in response to the escalating US-China tech war. The report mentions that its internal chip division was transferred to a Singaporean subsidiary called "Picoheart" in September. Some Chinese employees were even asked if they were interested in transferring to Singapore, a move seen as a risk-averse strategy to move sensitive technologies and operations out of geopolitical hotspots.

Analysis: A two-pronged strategy to build a strong computing power moat.

In my opinion, ByteDance's strategy is very pragmatic and aggressive.

On the one hand, it's willing to pay a premium or even lower prices to acquire NVIDIA, because this is currently the fastest and most mature path for training large models. The $140 billion budget demonstrates its determination not to fall behind in the model race.

On the other hand, while the self-developed chip's ability to "catch up with H20" may not sound as impressive as H100/H200, it is an important safety net. If the US completely bans even H20-level products, ByteDance will at least have its own chips to withstand the huge demand of its inference business and maintain the operation of TikTok and Doubao.

This hybrid strategy of "buying high-end products and developing mid-to-low-end ones," coupled with the shift of R&D focus to Singapore, shows that ByteDance is trying to build the most solid moat for its AI empire in a turbulent situation.

This is an advertisement.
Tags: AIByteDanceH20H200HBMNvidiaTokenArtificial wisdomByteDance
ShareTweetShare
Mash Yang

Mash Yang

Founder and editor of mashdigi.com, and student of technology journalism.

Post a responseCancel Reply

This site uses Akismet service to reduce spam.Learn more about how Akismet processes website visitor comments.

Translation (Tanslate)

Recent updates:

Say goodbye to expensive dedicated graphics cards! Intel Core Ultra Series 3 becomes the new standard for edge AI robot computing.

Say goodbye to expensive dedicated graphics cards! Intel Core Ultra Series 3 becomes the new standard for edge AI robot computing.

2026-05-21
AMD takes the lead in the industry! AMD announces mass production of its sixth-generation EPYC processor "Venice," manufactured using TSMC's 2nm process.

AMD takes the lead in the industry! AMD announces mass production of its sixth-generation EPYC processor "Venice," manufactured using TSMC's 2nm process.

2026-05-21
Facing formidable Asian rivals, Meta AI smart glasses enter the Japanese market, emphasizing a screenless design and real-time translation capabilities.

Facing formidable Asian rivals, Meta AI smart glasses enter the Japanese market, emphasizing a screenless design and real-time translation capabilities.

2026-05-21
mashdigi-Technology, new products, interesting news, trends

Copyright © 2017 mashdigi.com

  • About mashdigi.com
  • Place ads
  • Contact mashdigi.com

Follow us

Welcome back!

Login to your account below

Forgotten Password?

Retrieve your password

Hãy nhập tên người dùng hoặc địa chỉ email để mở mật khẩu

Log In
×

You are about to be redirected to an external website.

The link you clicked will open an external webpage:

In reciprocal calculation...
×

Want to take a break? We recommend the following content:

  • • Foxconn-affiliated Hon Hai Precision Industry Co., Ltd. invests $1.74 million to build a factory in Kentucky, but is it not for Apple?
  • • In the future, iOS platform app content will only be manageable on iPhone and iPad devices.
  • • Toyota unveils new autonomous driving test vehicle "TRI-P4" based on Lexus LS 500h

You can return by swiping the page or clicking anywhere.

No Result
View All Result
  • About mashdigi.com
  • Place ads
  • Contact mashdigi.com

Copyright © 2017 mashdigi.com