Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

Oracle cancels $7 billion Tencent chip lease as memory costs bite

Oracle cancelled a $7 billion chip leasing deal with Tencent after US officials raised concerns about it, Bloomberg reported on 4 November, citing the Financial Times.

TechnologyAnalysisRachel NwosuPublished: 4 November 20263 min readSources 21
Oracle cancels $7 billion Tencent chip lease as memory costs bite

Oracle pulls a $7 billion lease

The Financial Times reported last month that Oracle had signed a five-year agreement. The deal leased 100,000 chips to Tencent. It was the Chinese firm's largest ever overseas deal. US officials told Oracle of their concerns about the deal, the FT said, citing two people familiar with the matter.

Tencent did not immediately respond to a request for comment. The cancellation matters beyond the two companies. It leaves Tencent without a signed supply deal. It also lands as memory prices rise for hardware buyers.

Neither Oracle nor Tencent has said what the cancellation costs.

Apple warns on memory costs

Apple chief executive John Ternus spoke on a conference call on Monday. He said memory costs in the fourth quarter would be higher than in the previous three months. "What we're seeing with memory pricing right now is unprecedented," Ternus said.

Apple raised iPad and MacBook prices in June, then iPhone prices in September. Ternus said those increases had "provided some offset" for the company. Revenue rose 12% to $114.5 billion from $102.5 billion. Earnings came in at $2.10 a share against $1.99 expected in a FactSet poll. Shares fell 2.75% in after-hours trading.

Apple posted stronger-than-expected revenue in its fiscal third quarter. Net income was $30.71 billion, up from $27.47 billion a year earlier. Apple said earnings included five cents a share from tariff refunds.

Huawei sends buyers to the grey market

Huawei is telling some customers to source memory for its AI chips themselves. The Information reported this, according to The Decoder. The company plans to launch the Ascend 960DT early next year. It says it will double the performance of its current chips.

US rules block direct purchases from Samsung and SK hynix. Those two firms control more than 80% of the global HBM market. Huawei's Chinese suppliers cannot produce enough HBM. Smuggled HBM3E memory has jumped from a few hundred dollars to around $4,000 per unit in just one year, according to a Shenzhen broker cited by The Decoder. Each AI chip needs six to eight units, so the memory alone can cost up to $32,000.

The black market also draws on older Nvidia GPUs stripped for parts and excess inventory from chip designers that scrapped projects or cut orders.

The loophole: cheap processors, expensive memory

A smuggler told The Information that his chip design firm builds cheap, low-performing processors. Tom's Hardware relayed that account. The processors sit below the US and South Korean export threshold.

Those boards are paired with HBM from Samsung and SK hynix. That memory is banned for export to China on its own. The rules look only at processor performance, so the attached memory travels with it. Boards land in Hong Kong and cross into Shenzhen in commuters' luggage. Workshops heat the boards to soften the solder and lift the memory chips off. Despite a high failure rate, the high price of HBM makes the procedure lucrative.

What to watch

Huawei's Ascend 960DT is due early next year with its memory supply unresolved. South Korea's customs authority had uncovered trade violations worth $560 million by May, already past the previous full-year record. A third of that total involved strategic goods. Those included chips exported without authorisation or with false declarations.

Comments 0

Sources

21
  1. 01Oracle Scraps $7 Billion Deal to Lease Chips to Tencent, FT SaysEN
  2. 02AI chip smugglers exploit trade control loophole to bring banned HBM from South Korea into ChinaEN
  3. 03Huawei tells customers to find their own AI chip memory, effectively outsourcing procurement to the black marketEN
  4. 04Apple warns of ‘unprecedented’ rise in memory chip pricesEN
  5. 05EcoFlow launches TITAN Series backup ecosystem with up to $2,199 savings, Lectric limited-edition XP4 750 e-bike deals, moreEN
  6. 06Bosch gives chips their own division, eyeing €5bn in sales by 2030EN
  7. 07Chip Industry Technical Paper Roundup: Nov. 3EN
  8. 08India plans indigenous GPU chips in 24 months, says MeitY Secretary S Krishnan on AI strategy - ET GovernmentEN
  9. 09Is There Room for Small-Scale OSATs in India’s Chip Packaging Boom?EN
  10. 10 AI model gets a 32-bit RISC-V CPU compiled directly into its weights, then runs DOOM on it EN
  11. 11Marvell Snaps Photonics Chipmaker Lumentum to Bolster AI ConnectivityEN
  12. 12Li Auto launches 2026 i6 with unchanged starting price, upgraded in-house chip and batteryEN
  13. 13China’s CXMT defends bet on risky new DRAM architecture to circumvent US chip curbsEN
  14. 14RISC-V Technology - Global Strategic Business Report 2026: Capitalize On 25.1% CAGR To Secure Share Of A $3 Billion Market By 2032 Track Sifive, Andes TechnologEN
  15. 15FuriosaAI’s 3.8 Trillion Won Valuation: State Funds, No Chip Allies - 사이언스모니터EN
  16. 16As AI cyber risks grow, Chinese chip expert urges shared ethics standards with US rivalsEN
  17. 17AI, fuelled by advanced chips, set to power semiconductor manufacturing: industry veteranEN
  18. 18Chinese AI chipmaker Iluvatar CoreX says new GPU matches Nvidia’s H200 as chip gap closesEN
  19. 19Oracle Cloud to Offer Google TPUs and New AI ChipsEN
  20. 20Huawei’s Tau Scaling Law roll-out reaches base stations in blow to US chip curbsEN
  21. 21South Korea urged to guard chip profitability as China’s CXMT and YMTC gain market shareEN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Rachel Nwosu

Rachel Nwosu

AI, models and technology

Rachel Nwosu covers AI, models and technology for FLASH24, working from public model documentation, benchmark releases and repository histories rather than press summaries, and she skips announcements that arrive without reproducible numbers. She checks training-data claims against dataset cards and reruns reported metrics where code is available. She spends much of her week interviewing researchers and engineers, tracking model launch calendars, and comparing vendor benchmarks with independent evaluations. Outside the desk she runs 3D printers, restores old computers, and tests how models learn from internet junk. She does not publish benchmark figures she cannot trace to a source.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.