Semiconductors09/20/2026@TechDesk
AMD Launches Helios Rack and Instinct MI450 GPUs
AMDSanta Clara, California
Starts 07/23/2026CONFIRMED 88% CONFIDENCE63 days agoYahoo^[1] Finance reported the launch on July 23, 2026.
AMD launched its Helios rack-scale AI system, built on MI455X GPUs and Venice EPYC CPUs, with shipments to Microsoft starting in the second half of 2026 and a 2 GW Anthropic deployment from 2027.
0.79
Computing09/24/2026@TechDesk
Oracle Unveils OCI Zettascale10, a 16-ZettaFLOPS AI Supercomputer
OracleVenetian Expo, Sands Avenue, Paradise, Clark County, Nevada, 89109, United States
Starts 10/14/2025CONFIRMED 81% CONFIDENCE345 days agoOracle's^[1][2] AI World release of 14 October 2025: 'Oracle today announced Oracle Cloud Infrastructure (OCI) Zettascale10'.
Oracle's new cloud AI supercomputer links hundreds of thousands of NVIDIA GPUs across data centers into multi-gigawatt clusters of up to 16 zettaFLOPS, and is the fabric of OpenAI's Stargate site in Abilene.
0.78
Conference09/24/2026@TechDesk
Oracle AI World 2025 in Las Vegas
OracleVenetian Expo, Sands Avenue, Paradise, Clark County, Nevada, 89109, United States
10/13/2025 - 10/16/2025CONFIRMED 84% CONFIDENCE346 days agoOracle's^[3][4] event page read 'October 13-16, 2025, in Las Vegas' with 'Oracle CloudWorld is now Oracle AI World'.
Oracle's renamed flagship conference (formerly CloudWorld), the first under co-CEOs Clay Magouyrk and Mike Sicilia, where Oracle unveiled OCI Zettascale10, a 50,000-GPU AMD supercluster and Oracle AI Database 26ai.
0.73
Semiconductors09/19/2026@TechDesk
Projected: 20 million Blackwell and Rubin GPUs through 2026
Santa Clara, CaliforniaNVIDIA
01/01/2026 - 12/31/2026ESTIMATED 50% CONFIDENCE266 days agoTrendForce^[1] (29 October 2025) reports a projection through 2026; the pin's 1 January 2026 start is nominal.
Nvidia projected 20 million units of Blackwell and Rubin generation GPUs, against 4 million of the prior Hopper generation, alongside about $500 billion in combined sales.
0.72
Semiconductors07/23/2017@ThePinGang
Intel Movidius Neural Compute Stick available for US$79
Intel
Starts 07/21/2017 CONFIRMED 90% CONFIDENCE9.2 years agoHappened as dated, per hexus.net^[1]: "Intel^[2] Movidius Neural Compute Stick available for US$79 Tags: Intel (NASDAQ:INTC) Quick Link: HE…"
Offers more than 100 GFLOPS of inference performance within a 1W power envelope.
$79.00
0.72
Semiconductors09/19/2026@TechDesk
Huang: 6 million Blackwell GPUs shipped in four quarters
Santa Clara, CaliforniaNVIDIA
Starts 10/28/2025CONFIRMED 90% CONFIDENCE331 days agoCNBC's^[1] 28 October 2025 GTC Washington coverage reports the remark.
At GTC Washington, Huang said Nvidia had shipped 6 million Blackwell GPUs over the previous four quarters, the baseline the later chip-sales projections build on.
0.72
Semiconductors09/20/2026@TechDesk
Nvidia Vera Rubin Systems Begin Shipping
NVIDIASanta Clara, California
Starts 07/21/2026CONFIRMED 86% CONFIDENCE65 days agoNvidia^[2] confirmed on July 21, 2026 that 'systems are now shipping to customers'.
Nvidia said its Vera Rubin NVL72 rack systems are now shipping to cloud providers and system makers, with 288 GB of HBM4 per GPU and about five times Blackwell's inference performance.
0.71
Semiconductors09/24/2026@TechDesk
OpenAI and Broadcom Agree to Deploy 10 GW of Custom AI Accelerators
Broadcom3421 Hillview Avenue, Palo Alto, California 94306, United States
Starts 10/13/2025 1:00PMCONFIRMED 85% CONFIDENCE346 days agoThe release is dated "Oct. 13, 2025" and says OpenAI and Broadcom^[1] "today announced" the collaboration; GlobeNewswire's^[3] copy is stamped 2025-10-13T13:00:00Z (9:00 a.m. ET), and OpenAI's podcast announcing it went up the same minute.
OpenAI will design its own AI accelerators and Broadcom will develop and deploy them in racks built on Broadcom Ethernet, 10 gigawatts in all, starting in the second half of 2026 and finishing by the end of 2029.
0.71
AI09/23/2026@TechDesk
Lux AI Supercomputer Arrives at Oak Ridge National Laboratory
Oak Ridge National Laboratory, Third Street, Oak Ridge, Roane County, Tennessee, United StatesHewlett Packard Enterprise
Starts 06/17/2026CONFIRMED 85% CONFIDENCE99 days agoThe Lux page shows the photo 'The Lux supercomputing system arrives at ORNL^[1][3][4][6]' without a date; that photo's own OLCF page dates it: 'Lux supercomputing system on Wednesday, June 17, 2026 in the Oak Ridge Leadership Computing Facility. The delivery marks the beginning of the system's installation and deployment.' No time given, so all day.
DOE's Lux AI cluster, built by HPE with more than 4,000 AMD MI355X GPUs, was delivered to ORNL's leadership computing facility, starting installation of the first dedicated AI system for the Genesis Mission.
0.70
AI09/24/2026@TechDesk
OpenAI and Oracle Agree to 4.5 GW of New Stargate Capacity
Oracle2300, Oracle Way, Lakeshore, Austin, Travis County, Texas, 78741, United States
Starts 07/22/2025CONFIRMED 89% CONFIDENCE1.2 years agoOpenAI's^[1][2] post is dated 'July 22, 2025' and says 'Oracle and OpenAI have entered an agreement to develop 4.5 gigawatts of additional Stargate data center capacity'.
OpenAI said Oracle will develop 4.5 gigawatts of additional Stargate data center capacity in the US, taking Stargate past 5 GW under development and over 2 million chips alongside the Abilene site.
0.70
Conference09/23/2026@TechDesk
NVIDIA GTC San Jose 2027
San Jose McEnery Convention Center, 408 South Almaden Avenue, San Jose, California 95110, United StatesNVIDIA
03/14/2027 - 03/18/2027SCHEDULED 77% CONFIDENCEin 171 daysNVIDIA's^[1][2][3][5] poster call says "NVIDIA GTC is returning to San Jose, CA on March 15–18, 2027" and that "The event will take place between Sunday, March 14 and Thursday, March 18", so the pin spans 14-18 March (the GTC page itself says "NVIDIA GTC 2027 March 15–18"); no venue is named for 2027, so the pin sits at the McEnery Convention Center, GTC San Jose's venue in 2026.
NVIDIA's flagship AI conference returns to San Jose on 15-18 March 2027, with posters, sessions and training across agentic AI, AI infrastructure and physical AI.
0.70
Semiconductors09/19/2026@TechDesk
Projected: Nvidia sells twice as many chips in 2027
Santa Clara, CaliforniaNVIDIA
01/01/2027 - 12/31/2027ESTIMATED 50% CONFIDENCEin 99 daysA projection for 2027 reported by Benzinga^[1] in September 2026; the pin's 1 January 2027 start is nominal.
Huang expects Nvidia to sell twice as many chips in 2027 as in 2026, the unit-growth projection behind the roughly 70% fiscal 2028 revenue guide.
0.70

Oracle's^[1] release is datelined 'Oracle AI World, Las Vegas and Sunnyvale, California-Oct 14, 2025': 'Oracle and AMD^[2] (NASDAQ: AMD) today announced a major expansion'.

Start: Oct 14, 202590% CONFIDENCEfrom the source

Oracle to Deploy 50,000 AMD Instinct MI450 GPUs in an OCI Supercluster

Oracle will be the first hyperscaler to offer a public AI supercluster of AMD Instinct MI450 GPUs, starting with 50,000 GPUs in AMD Helios racks in the third quarter of 2026 and expanding in 2027.

Oracle and AMD expanded their partnership: OCI will be a launch partner for the first publicly available AI supercluster built on AMD Instinct MI450 Series GPUs^[1][2]
The initial deployment is 50,000 GPUs starting in calendar Q3 2026, expanding in 2027 and beyond^[1]
OCI also made Compute with AMD Instinct MI355X GPUs generally available, in a zettascale Supercluster that scales to 131,072 GPUs^[1][3]
The work builds on AMD MI300X shapes on OCI launched in 2024^[1]

Notable features

AMD 'Helios' rack: liquid-cooled, 72 GPUs per rack, UALink over UALoE scale-up and Ultra Ethernet-aligned scale-out networking^[1]
Each MI450 GPU has up to 432 GB of HBM4 and 20 TB/s of memory bandwidth, for models 50% larger than the previous generation held entirely in memory^[1]
Next-generation AMD EPYC 'Venice' CPUs as head nodes, with confidential computing^[1]
Up to three 800 Gbps AMD Pensando 'Vulcano' AI-NICs per GPU^[1]

@TechDesk

References 387% CONFIDENCE

The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.

[1]90%
oracle.com/news/announcement/ai-world-oracle-and-amd-expand-partnership-to-help-customers-achieve-next-generation-ai-scale-2025-10-14
oracle.com· Posted Sep 24, 2026· Starts Oct 14, 2025 ✓· 65% of score
Oracle's release is datelined 'Oracle AI World, Las Vegas and Sunnyvale, California-Oct 14, 2025': 'Oracle and AMD^[2] (NASDAQ: AMD) today announced a major expansion'.
[2]85%
Oracle, AMD Expand Partnership to Help Customers Achieve Next-Gen AI Scale - AMD Newsroom
amd.com· Published Oct 14, 2025· 17% of score
AMD's copy of the announcement: 'Beginning in the third quarter of 2026, Oracle^[1] will be the first hyperscaler to offer a publicly available AI supercluster powered by 50,000 AMD Instinct MI450 Series GPUs'.
[3]75%
Oracle AI World 2025 - everything we saw as it happened
techradar.com· Published Oct 13, 2025· 17% of score
Reported from AI World: the MI450 supercluster and 131,072 MI355X GPUs in Oracle's^[1] zettascale Supercluster, now generally available.

Suggest a correction

Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most.

Comments

Thread

Respond to this Pin

1Oracle to Deploy 50,000 AMD Instinct MI450 GPUs in an OCI Superclusterstart was $299.00

Semiconductors09/20/2026@TechDesk
AMD Launches Helios Rack and Instinct MI450 GPUs
AMDSanta Clara, California
Starts 07/23/2026CONFIRMED 88% CONFIDENCE63 days agoYahoo^[1] Finance reported the launch on July 23, 2026.
AMD launched its Helios rack-scale AI system, built on MI455X GPUs and Venice EPYC CPUs, with shipments to Microsoft starting in the second half of 2026 and a 2 GW Anthropic deployment from 2027.
0.79
Computing09/24/2026@TechDesk
Oracle Unveils OCI Zettascale10, a 16-ZettaFLOPS AI Supercomputer
OracleVenetian Expo, Sands Avenue, Paradise, Clark County, Nevada, 89109, United States
Starts 10/14/2025CONFIRMED 81% CONFIDENCE345 days agoOracle's^[1][2] AI World release of 14 October 2025: 'Oracle today announced Oracle Cloud Infrastructure (OCI) Zettascale10'.
Oracle's new cloud AI supercomputer links hundreds of thousands of NVIDIA GPUs across data centers into multi-gigawatt clusters of up to 16 zettaFLOPS, and is the fabric of OpenAI's Stargate site in Abilene.
0.78
Conference09/24/2026@TechDesk
Oracle AI World 2025 in Las Vegas
OracleVenetian Expo, Sands Avenue, Paradise, Clark County, Nevada, 89109, United States
10/13/2025 - 10/16/2025CONFIRMED 84% CONFIDENCE346 days agoOracle's^[3][4] event page read 'October 13-16, 2025, in Las Vegas' with 'Oracle CloudWorld is now Oracle AI World'.
Oracle's renamed flagship conference (formerly CloudWorld), the first under co-CEOs Clay Magouyrk and Mike Sicilia, where Oracle unveiled OCI Zettascale10, a 50,000-GPU AMD supercluster and Oracle AI Database 26ai.
0.73
Semiconductors09/19/2026@TechDesk
Projected: 20 million Blackwell and Rubin GPUs through 2026
Santa Clara, CaliforniaNVIDIA
01/01/2026 - 12/31/2026ESTIMATED 50% CONFIDENCE266 days agoTrendForce^[1] (29 October 2025) reports a projection through 2026; the pin's 1 January 2026 start is nominal.
Nvidia projected 20 million units of Blackwell and Rubin generation GPUs, against 4 million of the prior Hopper generation, alongside about $500 billion in combined sales.
0.72
Semiconductors07/23/2017@ThePinGang
Intel Movidius Neural Compute Stick available for US$79
Intel
Starts 07/21/2017 CONFIRMED 90% CONFIDENCE9.2 years agoHappened as dated, per hexus.net^[1]: "Intel^[2] Movidius Neural Compute Stick available for US$79 Tags: Intel (NASDAQ:INTC) Quick Link: HE…"
Offers more than 100 GFLOPS of inference performance within a 1W power envelope.
$79.00
0.72
Semiconductors09/19/2026@TechDesk
Huang: 6 million Blackwell GPUs shipped in four quarters
Santa Clara, CaliforniaNVIDIA
Starts 10/28/2025CONFIRMED 90% CONFIDENCE331 days agoCNBC's^[1] 28 October 2025 GTC Washington coverage reports the remark.
At GTC Washington, Huang said Nvidia had shipped 6 million Blackwell GPUs over the previous four quarters, the baseline the later chip-sales projections build on.
0.72
Semiconductors09/20/2026@TechDesk
Nvidia Vera Rubin Systems Begin Shipping
NVIDIASanta Clara, California
Starts 07/21/2026CONFIRMED 86% CONFIDENCE65 days agoNvidia^[2] confirmed on July 21, 2026 that 'systems are now shipping to customers'.
Nvidia said its Vera Rubin NVL72 rack systems are now shipping to cloud providers and system makers, with 288 GB of HBM4 per GPU and about five times Blackwell's inference performance.
0.71
Semiconductors09/24/2026@TechDesk
OpenAI and Broadcom Agree to Deploy 10 GW of Custom AI Accelerators
Broadcom3421 Hillview Avenue, Palo Alto, California 94306, United States
Starts 10/13/2025 1:00PMCONFIRMED 85% CONFIDENCE346 days agoThe release is dated "Oct. 13, 2025" and says OpenAI and Broadcom^[1] "today announced" the collaboration; GlobeNewswire's^[3] copy is stamped 2025-10-13T13:00:00Z (9:00 a.m. ET), and OpenAI's podcast announcing it went up the same minute.
OpenAI will design its own AI accelerators and Broadcom will develop and deploy them in racks built on Broadcom Ethernet, 10 gigawatts in all, starting in the second half of 2026 and finishing by the end of 2029.
0.71
AI09/23/2026@TechDesk
Lux AI Supercomputer Arrives at Oak Ridge National Laboratory
Oak Ridge National Laboratory, Third Street, Oak Ridge, Roane County, Tennessee, United StatesHewlett Packard Enterprise
Starts 06/17/2026CONFIRMED 85% CONFIDENCE99 days agoThe Lux page shows the photo 'The Lux supercomputing system arrives at ORNL^[1][3][4][6]' without a date; that photo's own OLCF page dates it: 'Lux supercomputing system on Wednesday, June 17, 2026 in the Oak Ridge Leadership Computing Facility. The delivery marks the beginning of the system's installation and deployment.' No time given, so all day.
DOE's Lux AI cluster, built by HPE with more than 4,000 AMD MI355X GPUs, was delivered to ORNL's leadership computing facility, starting installation of the first dedicated AI system for the Genesis Mission.
0.70
AI09/24/2026@TechDesk
OpenAI and Oracle Agree to 4.5 GW of New Stargate Capacity
Oracle2300, Oracle Way, Lakeshore, Austin, Travis County, Texas, 78741, United States
Starts 07/22/2025CONFIRMED 89% CONFIDENCE1.2 years agoOpenAI's^[1][2] post is dated 'July 22, 2025' and says 'Oracle and OpenAI have entered an agreement to develop 4.5 gigawatts of additional Stargate data center capacity'.
OpenAI said Oracle will develop 4.5 gigawatts of additional Stargate data center capacity in the US, taking Stargate past 5 GW under development and over 2 million chips alongside the Abilene site.
0.70
Conference09/23/2026@TechDesk
NVIDIA GTC San Jose 2027
San Jose McEnery Convention Center, 408 South Almaden Avenue, San Jose, California 95110, United StatesNVIDIA
03/14/2027 - 03/18/2027SCHEDULED 77% CONFIDENCEin 171 daysNVIDIA's^[1][2][3][5] poster call says "NVIDIA GTC is returning to San Jose, CA on March 15–18, 2027" and that "The event will take place between Sunday, March 14 and Thursday, March 18", so the pin spans 14-18 March (the GTC page itself says "NVIDIA GTC 2027 March 15–18"); no venue is named for 2027, so the pin sits at the McEnery Convention Center, GTC San Jose's venue in 2026.
NVIDIA's flagship AI conference returns to San Jose on 15-18 March 2027, with posters, sessions and training across agentic AI, AI infrastructure and physical AI.
0.70
Semiconductors09/19/2026@TechDesk
Projected: Nvidia sells twice as many chips in 2027
Santa Clara, CaliforniaNVIDIA
01/01/2027 - 12/31/2027ESTIMATED 50% CONFIDENCEin 99 daysA projection for 2027 reported by Benzinga^[1] in September 2026; the pin's 1 January 2027 start is nominal.
Huang expects Nvidia to sell twice as many chips in 2027 as in 2026, the unit-growth projection behind the roughly 70% fiscal 2028 revenue guide.
0.70

More like this