Robert Brooks IV
@boborado
Followers
3K
Following
8K
Media
176
Statuses
2K
founding team @lambdaapi
SF & SF (South Florida)
Joined October 2012
🚨 Hyperscalers are running low on GPUs. At @LambdaAPI — we have 2 simple solutions:. 1. Our cloud has thousands of available GPUs from A100 to A10 and soon H100 ☁️ . 2. We can build you the exact ‘cloud’ configuration you want in our datacenter with an OpEx model payment plan 💰.
8
90
978
I have 248 H100 SXM5s networked with 3200 Gbps Infiniband just sitting in front of me. Ready to go. Who wants to do something out of this world cool? Let’s train your foundation model or LLM. DMs open.
56
55
453
Yes, it’s true. If you’re just discovering Lambda, all we do is build large scale GPU clusters for ML in the cloud: For this specific initiative — we’re looking to partner with a deep learning team training a large foundation model. Whether you’re an.
7
16
143
when you're deciding which founder to invest in, ask yourself: do they have a tattoo of their company logo on their arm muscles?
11
4
124
Took thousands of 8x-H100s and chunked them up into 4x, 2x & 1x instances. Lower price points + more options for the AI community! Enjoy 🤠
4
3
123
By far one of the coolest things was Jensen starting a “Lambda, lambda, lambda, lambda, lambda” chant
8
10
99
Train train. GPUs so good. GPUs so good. Model weights go pop pop pop.
2
8
87
Announced today ✅. Launched today ✅. NVIDIA GH200 on Lambda Cloud starting at $5.99/hour for *576GB of total addressable VRAM*. Compared to H100 on Lambda Cloud, a comparable amount of VRAM (80GB x 8 H100s = 640GB) costs $20.72/hour.
Lambda Cloud Clusters are now available with the new NVIDIA GH200 Grace Hopper™ Superchip.
1
8
66
Did you hear it?. Lambda was mentioned in NVIDIA’s opening remarks during their record earnings call today 😁. Minute 4:
7
4
53
Get a weather service that looks at you the way @NWSBayArea looks @fog_karla.
@fog_karla Technically, no, you're condensation, but it is in your heart to be who you are and we love you for it. Know that all times you're just one step from precipitation, really just one step. Do not to fret, for so many great things in life it is the journey, not the destination. 🌧️.
1
7
44
$140k/month?. You can rent 2x 8-H100 nodes connected with 3.2Tb/s Infiniband from @LambdaAPI for $48k/month. Available on-demand, no long term contract required:
400 billion parameters.6.4 trillion bits (half precision).800GB.A H100 has 80GB.A H100 node has 8H100s, so it needs 2 nodes.$140k monthly bill🫢💸.$70k with 8-bit quantization. I hope people do good distillation or extreme non-degrading quantization (or that I did my math wrong).
2
3
46
@sergeykarayev Because hyperscalers are their largest customers by several orders of magnitude. DGX Cloud is a step in that direction. Yet, that product is still hosted by… hyperscalers. NVIDIA, for all the grief it gets, is actually insanely partner friendly and only sells indirect.
2
2
43
Want to fine-tune Llama3 on 64-512 GPUs next week? DM me.
6
8
44
Will be giving out 10^26 in GPU cloud credits to AI startups as soon as this bill is defeated.
3
1
39
Best wedding surprise!?. My mother in law went the extra mile and got me a @Chiptos_nft wedding rehearsal cake! 🤣🤖🌴
9
6
38
*Update*. To all that have reached out, thank you 🙏🏻. Here’s the process we’re running:. - Launch an official program to secure everyone’s interest tomorrow. - Meet w/ teams that are serious about an OS foundation model. - Announce a partner on 5/31. - Release model in July ✨.
3
2
32
@dylan522p Yes. Unique to the H100. It’s 32 nodes but 1 node is dedicated to the network among other orchestration tasks for the cluster. Writing a blog post on this now.
1
2
34
I could say so much but the most important thing is THANK YOU to our customers. We exist to serve AI teams!.
Lambda is excited to announce we have raised a $320M Series C, achieving a $1.5B valuation. This new financing will be used to expand our AI cloud and further our mission to build the #1 AI compute platform in the world.
1
0
32
We're not talking about DGX GH200 enough. Per Jensen: "It is one giant GPU". *144 TB GPU Memory*. Who wants one? I'd like to speak with you as I'm. working on something special ✨
8
1
32
👋🏻 big NeRF guy here. Had a little too much fun with @LumaLabsAI iOS App today . Do you folks realize you can create NeRFs with your iPhone!?!? It trains the neural net in real time!. So much fun. Here’s my first one. Plenty more to come.
2
3
30
This structure is unique in that Lambda’s on-demand customer base is so large and so diverse that we’ve pioneered a debt raise solely based on customer mix (thousands of AI companies), *not* customer concentration ❤️.
Today, we are announcing $500 million in debt financing to scale Lambda’s On-Demand GPU Cloud. Learn more:
1
2
29
Lambda’s HQ is 6 minutes from NVIDIA HQ. Who wants a factory tour next week? DMs open.
2
2
28
Nothing like NeurIPS! Had this conversation 100x so wanted to post here:. YES! NVIDIA selected Lambda Cloud as a launch partner for both GH200 & H200: . - . - GH200 is live with customers TODAY:
2
1
29
Honestly I'm most excited for @abhi_venigalla today. That dude rocks. But seriously, congrats to the entire @MosaicML team. Absolutely wow!.
2
0
28
1
0
28
More H100s & A10s available on Lambda's cloud. Happy training!.
1
1
26
‼️‼️Lambda has solved something only a couple of other (very large) companies have: partitioning a large Infiniband deployment (thousands of H100s) to make smaller virtualized clusters . However, only Lambda makes these self-serve (no humans in the loop!) and. on-demand! Spin.
Introducing Lambda 1-Click Clusters: On-Demand GPU Clusters featuring NVIDIA H100 Tensor Core GPUs with Quantum-2 InfiniBand. No long-term contract required.
1
7
27
A $100M Cluster for Nat’s AI startup portfolio is crazy cool👏🏻 . Lambda’s $20M Cluster for building an open-source foundation model to give back to the ML community is a tad cooler 😉 . Either way — so awesome to see the largest barrier to entry for AI startups be brought down!.
1
2
24
Melt GPUs on the regular? Come over to Lambda + get a free hat 🎩.
This drip was brought to you by an A100 instance I left running for like a week when LLaMA came out . Thanks @boborado and @LambdaAPI!
5
1
24
Super excited to announce our partnership with @weights_biases today. Two companies that are 100% dedicated to serving the Machine Learning Engineer!.
We’re excited to roll out a new partnership with @LambdaAPI to accelerate deep learning and #AI initiatives, unlocking new use cases across industries.
2
4
24
Life update📢: After an amazing decade at Google/DeepMind, I’m thrilled to announce that I’ll be joining @OpenAI in a few weeks! I’m excited for the opportunity to co-build the OpenAI Zürich office alongside my close collaborators @giffmana and.@__kolesnikov__ 🚀.
1
1
23
NVIDIA H200 MLPerf Results Today ✅. H200 pricing live in Lambda Cloud Today ✅.
What do NVIDIA Hopper GPUs with 141GB of HBM3e get you? NVIDIA H200 Tensor Core GPUs delivered up to 1.5X faster Llama 2 70B inference test in MLPerf compared to H100. Our H200 pricing is now LIVE: Read new NVIDIA benchmark blog:
1
1
23
Calling all MLEs — slide into my DMs!. I’ll hook up cloud credits for feedback on a very exciting upcoming product launch ☁️.
1
5
21
1
5
22
👀 . "Lambda made 9-figures $ worth of H100s available in its on-demand cloud and sold out in just over an hour".
🪩The @stateofaireport 2023 is now here. Our 6th installment is one of the most exciting years I can remember. The #stateofai report covers everything you *need* to know, covering research, industry, safety and politics. There’s lots in there, so here’s my director’s cut 🧵
2
2
22
NVIDIA's first ARM CPU is now available on-demand ;).
Just Launched 🎉 - NVIDIA GH200 Superchip on Lambda On-Demand! Spin up your own instance today and enjoy up to 2x faster time-to-first-token (TTFT) with models like Llama3 70B. Try it out today 👉
1
1
22
@HPC_Guru Looking for a Head of High Performance Computing to build a team designing clusters for deep learning (remote). We already have multiple 64-800+ GPU deployments at major National Labs & the Fortune 500. Anyone come to mind?. cc: @hpcprogrammer @hpcjoe @hpcgal @hpcnotes.
1
12
21
More GPUs incoming this week.
Was loving the pricing on @LambdaAPI Cloud, but apparently everyone else does too. Was good while it lasted 🙃
1
1
19
If you’ve worked w/ Lambda in the past 5 years, chances are you’ve been on a call where @mitesh711 and I finish each others sentences. Now in podcast form!. We’re beyond excited to share our journey making a 248x GPU cluster available to train an open-source foundation model!.
New Open Source Startup Podcast 🎙️. Why should foundation models be open source? . How is @LambdaAPI helping make it a reality? . Listen in on the latest convo from @amanda_robs & @tnachen with Lambda founding team members @boborado & Mitesh to find out 👇.
1
7
20
Are there companies that still train models from scratch (not named Meta, OAI, MSFT & GOOG)?. Hope pretraining has not been left for dead. That would be the bitterest of lessons.
12
2
21
On a mission to ensure every ML team is GPU rich! . Special thanks to Lambda customer @gorkemyurt for the swag hook up!
2
3
21
We just crossed over 💯 submissions!. Foundation models that span 3D, video, radiology, drug discovery, geospatial, legal and healthcare ✨. I’m already on my 2nd Throat Coat Tea as I’ve met with dozens in the last 72 hours 😆. Let’s do something special for the ML community!.
I have 248 H100 SXM5s networked with 3200 Gbps Infiniband just sitting in front of me. Ready to go. Who wants to do something out of this world cool? Let’s train your foundation model or LLM. DMs open.
4
5
20
Played with @googlecardboard all weekend. My first thought? I should start a company for custom sales pitches/product demos thru VR ✅.
3
0
16
Lambda just announced 1 & 2 year H100 contracts as well as 0% down payment. Available Q2 2024.
4
2
19
I need to order more Fal swag.
Lambda reserved support was already top-notch, but seeing on-demand support to get onto that level is amazing, ticket to resolution under 20 minutes! Well done @LambdaAPI @boborado
0
2
19
The worst-kept secret in the ML community is thousands of AI teams already used Lambda for Inference, despite not having a dedicated Inference API and TaaS (Tokens-as-a-Service) offering. Today, 150,000+ ML Engineers & Researchers are empowered with a dedicated platform for
3
3
19
Um, I live less than 2 hours from @NASAKennedy. Got to work on a cool project for @blueorigin during my last gig in 2018. Will be watching New Glenn’s maiden flight next week in person! . 320 feet tall, one of the largest vehicles ever built. Helluva way to start 2025!
1
2
18
2024 AI predictions:. - new NN architecture dramatically reduces the # of accelerators needed to train models.- humanoids and “tangible AI” become a bigger deal.- foundation models in specific domains have major breakthrough: geospatial, healthcare, robotics, etc.- OpenAI starts.
1
4
16
@levelsio @marckohlbrugge @dannypostmaa Yes & no. In addition to cloud, Lambda sells on-prem infra that can be put in our colo and we manage everything for you. Significantly bringing down costs while also abstracting away the pains. Happy to help here.
1
0
17
@stephenbalaban Can I look outside the Silicon Valley echo chamber through an Apple Vision Pro?.
1
0
17
Holy s*** I cannot wait to show y'all what's cooking.
I have 248 H100 SXM5s networked with 3200 Gbps Infiniband just sitting in front of me. Ready to go. Who wants to do something out of this world cool? Let’s train your foundation model or LLM. DMs open.
1
2
17
I’m getting married in 2 days. Here is my absolute best unintended hack for staying relaxed during your wedding week:. Get wildly engrossed in a good sci-fi book. I’ve now barreled through 400 pages of the first Dune book in 3 days. So. Damn. Good. So. Damn. Relaxed.
5
0
14
Technical Founders,. You never need to read a marketing-y blog post about OKRs again. "An OKR is, effectively, you shouting out your intentions to the universe.". More here:
2
3
15
Thinking in decades is a superpower. Ability to live in multiple locations is also a superpower. Which leads me to this conclusion:. “remote forever companies” just unlocked a supercycle of positive change for the MOST ambitious of us in the human race. Allow me to explain:.
2
1
13
Little birdie told me there may be a lot more A10s available on our on-demand cloud. .
1
3
14
Problem: ML teams don’t like getting locked into multi-year contracts with the hyperscalers. Solution: ML teams can now rent massive GPU clusters on-demand, for weeks at a time! . *Perfect for a quick fine-tune of Llama 3.1 405B!* 👌.
0
3
15
James is one of the smartest people I’ve ever had the chance to work alongside. Must read:.
I've been watching markets compulsively for the last few years, and it's obvious that we're headed for some very strange times. I've been meaning to write about this - and how Bitcoin ties in - for a while, but finally got around to it this weekend.
1
1
14
Tokyo continues to blow my mind:. I ate at a Michelin star restaurant for lunch yesterday. Didn’t even know it. $13 for the best Udon I’ve ever had.
1
1
13
Saw a guy at a coffee shop training an LLM from scratch on proprietary data with his own cloud cluster. No open source datasets. No fine tuning. No free VC compute. Like a psychopath.
1
0
13
@chrisalbon If you care about H100, Lambda will be the first cloud with on-demand H100 before end of month 🚂🚂🚂.
0
0
14
Hard to explain to folks younger in their career that you don’t get to make strategic decisions and tweak the dials at a high level until you become a master of the details and execution. It can take a decade. That’s ok! Lean in!.
1
0
13
Probably the cutest thing in the world is my grandmother-in-law, who lives north of Miami, cuts out crypto newspaper articles and mails them to me in California 😭😌
1
0
12
Waymo doing 100,000 trips/week. Wow. Phenomenal presentation from @nathanbenaich right now at the official 2024 State of AI launch!
2
0
10
VC: “What’s your guys moat”. Me: “I have way better music taste than their entire founding team”.
4
0
12
1
1
12
Senior Software Dev in 2023: “I got really into object oriented programming and objective-c in the 2010s and then basic standard English in the 2020s”.
This is awesome - you can program your own personalized assistant in. English. This hottest programming language is also older than any other by several hundred years. And now you can execute it with general-purpose text-based computers.
0
0
10
“Ok, let’s do it” is the most underrated powerful sentence in business.
3
1
11
Who wants to get dinner at NeurIPS? I have reservations at Arnauds & Antoine’s 😊😊😊.
2
0
11
🙌🏻🙌🏻🙌🏻🙌🏻🙌🏻🙌🏻🙌🏻🙌🏻🙌🏻🙌🏻🙌🏻🙌🏻
In 2019, OpenAI announced GPT-2 with this post:. Today (~5 years later) you can train your own for ~$672, running on one 8XH100 GPU node for 24 hours. Our latest llm.c post gives the walkthrough in some detail:. Incredibly, the
0
1
11
Yeehaw!.
NVIDIA HGX H100 systems are now available on Lambda Cloud with 8x H100 SXM GPU instances on-demand for only $2.59/hr/GPU. Launch an instance while they last: Read the full announcement:
2
0
11
By far the greatest company perk is paying for your employees chatGPT Plus & MidJourney Pro subscriptions.
4
1
11