Aman Singla

Cerebras is bringing its AI supercomputers to the UAE’s Stargate project ◾ Deploying megawatt-scale AI infrastructure across the Gulf, India & Pakistan ◾ Partnering with Abu Dhabi’s G42 ... part of the U.S.-UAE Stargate deal ◾ Competing head-on with Nvidia in AI compute ◾ Planning 12–15 data centers within 8 months ◾ IPO refiling “soon,” says CEO Andrew Feldman The UAE is quietly building the world’s largest AI hub outside the U.S.

7

Oru'el's primary traction came in December when I met Pratyush Kumar of Sarvam at an event. "More compute without adding more GPUs, how?" he asked and we strolled off the event for 20 odd minutes. I explained to him what happens BTS, how does 'predciting' GPU failures affect this and how Foundational AI companies, and ultimately users, suffer. The sad part is, this is not something anyone could have controlled. It's really hard to predict as well. Grey failures is one of the biggest problems with GPUs and as quoted by Meta, 58% of their training run failures during Llama 3 were GPU related! Last week, Ajith and Nihith finalised the models and the architecture for our MVP. It outperforms what Bytedance did by a margin. Tho, there's no formal benchmark that exists, we would soon make one. Every SRE we have talked to told us the same 3 things : - We don't know when it would fail - We don't know why it would fail - We don't know what is the cause of failure until thorough post-mortem And since engineers don't really care about CapEx, the billions lost monthly across the world is just another headache. Oru'el's system works on pinpointing failures, telling you particular GPU about to fail, the cause of it and how much time is left to act with an actionable that they can swiftly follow. Latency? sub-milisecond. This is an SRE's dream come true. Predicting GPU failures be it for training or inference not only optimises the work flow but also calls for maintenance -- not on scheduled calendar but when you need it. This also tells you the 'health' of a GPU in real time, treating Silicon like a living being. Oru'el will be piloting soon and we have got our first prospects to work with. Everytime we cold called and talked about this to the decision makers, we never got a 'no' and that's a sign you know this industry needs physics to solve a physics problem -- not RL slapped on stats. oru-el.com

44

1 Comment

Last night Ashish Bhatia gave a demo session to 250 (of 420 regs) at our monthly GAI Insights "Learning Lab" Zoom call on Claude Code and OpenClaw For Claude Code, his 3 key takeaways are: - Anyone can Build: Technical/Non-Technical, any one can build. If you have an idea you can bring it to life. - Mins -> Hours -> Days: AI Tools can perform long horizon tasks. - Gulf of Specification: How do you convey your intent most effectively with AI models/agents. His 3 takeways for OpenClaw are: - Orchestration and not the model: Models are fungible and intelligence is getting commoditize. - AI is becoming the Operating system: Transition from answers to actions. - Agent Manager mindset: No upper limit to resources, what tasks would you offload to your agents. Link to recording in the comments. Let's keep Learning and Building!

61

2 Comments

Zero funding. Zero plan. Zero IP. One original product idea.  But your buddy still expects cofounder-level equity. Most cap table disasters start with guilt, not greed: “Am I ripping off my "cofounder" if I offer 1-3% equity?” Someone asked me this question years ago. Here’s their situation and what I advised: - Friend had the original product idea. - Sees himself as advisor or early employee. - Brings no funding, no plan, no intellectual property. Just the idea. The founder is asking: "Do I owe him equity?" Here's what I told him: Having the idea doesn't make you a cofounder. Ideas are cheap. Execution is everything. If he didn't bring: - A plan to execute - Funding to start - Intellectual property to build Then you don't "owe" him anything. But here's the relationship reality: Do the right thing anyway. Ask him what he thinks he should receive. Remind him that ideas alone don't translate to execution vision. If he asks for something reasonable,  give it to him. Why? Because every point of equity you give now  prevents future relationship disasters. 1-2% equity as goodwill costs you nothing today. But relationship resentment can costs you a lot down the road. Even if this person isn’t your close friend.  Reputational harm and resentment can travel in interesting ways. Be mindful how you communicate. ”Just business” never feels like “just business.” Especially amongst friends and family. Personalities are tricky things.  To their credit, angels and VCs understand that throwing  a couple points at the idea person. Especially if he's contributing some time and money. Even more if he's your first product tester. The real question is: "What's this relationship worth?" Ideas are worthless. Execution partnerships are priceless. If he transitions into actual contribution,  that 1-3% becomes a good investment, potentially. If he doesn't, it's cheap insurance against drama. Protect the relationship. Give the equity. Now most of you will say 1% - 3%?  That is insane for a co-founder title. Maybe. And…most friends and family will not understand  the nature of startup percentages (1%?! why not 15%?). Good point. Also likely true. And…if you’re raising VC money  won’t a 3% co-founder raise massive red flags? Again, solid point. All three of the above factors probably warrant   an entire post worth of dissection. Though in this specific high level scenario,  I’d say 1%-3% makes sense. What do you think?

22

10 Comments

Reading Metronome’s Monetization Operating Model, I kept coming back to one idea: pricing has become product. Software now delivers outcomes, not access. Yet most companies still charge as if they’re selling seats or licenses. That disconnect creates friction: for customers, unpredictability; for companies, stalled growth. The paper’s argument is simple but sharp—monetization isn’t a late-stage decision. It’s strategic infrastructure. Pricing needs the same ownership and iteration as any feature. Treat it like a surface that customers touch, not a spreadsheet buried in finance. If value is continuous and dynamic, pricing must be as well. That means product, GTM, finance, and engineering working from one system of truth. How many of us still treat pricing as an afterthought—when it should be a growth engine? https://lnkd.in/gnH7WzYf #Monetization #ProductStrategy #AI

8

As part of the IndiaAI Mission, Avataar is building UAI (Unified AI Interface) - a foundational DPI layer with thin, durable rails for trust and interoperability, connecting models, data and compute into a cohesive, discoverable ecosystem. We are privileged to collaborate with Prof. Ramesh Raskar (MIT) and Project NANDA: Architecting the Internet of AI Agents in shaping a broader “Internet of AI Agents” vision - exploring how a discovery layer for India’s evolving digital public infrastructure can unlock openness, innovation and responsible scale together. Equally grateful to Aafreen Siddiqui Sherwani (United Nations) for anchoring the dialogue, and to Pradeep Gupta (NVIDIA), Tanvi Lall (EkStep Foundation), Vineet Singh (Digital Green) and Jibu Elias (Mozilla Foundation) - for a thoughtful, forward-looking exchange on what it truly takes to build open AI platforms for India. The real opportunity ahead is not just technical, it is architectural. How we design these rails now will shape how intelligence compounds across citizens, enterprises and governance systems for decades. What principles should guide the design of AI infrastructure at national scale? #IndiaAI #DigitalPublicInfrastructure #ViksitBharat

180

9 Comments

Akash Network Q3 2025 Report - Overclock Labs, creators of Akash Network is the world's first distributed peer-to-peer marketplace for #AI cloud compute: ✅ Fee revenue increased 11% QoQ to 715,000 AKT, and rose 4% to $860,000, supported by higher deployment volume and incremental improvements to billing and operational tooling. ✅ New leases grew 42% QoQ to 27,000, reflecting increased model availability, mid-quarter Console and API enhancements, and greater developer activity ahead of upcoming GPU expansion initiatives. ✅ GPU demand remained stable, with usage at 367 units (-1% QoQ) and utilization consistently above 50%, indicating steady consumption of decentralized inference workloads despite a reduction in total capacity. ✅ Governance activity continued to progress, supporting community programs, analytics infrastructure, marketing initiatives, and provider incentive pilots in preparation for upcoming GPU expansion. ✅ Akash expanded its #AI model support with the integration of GPT-OSS-120B, Qwen3-Next-80B-A3B, Pluralis Node-0-7.5B, and DeepSeek-V3.1, while Console and API upgrades improved usability and contributed to increased tenant experimentation. Great progress, thanks team Greg Osuri, Cheng Wang Thanks Messari for publishing this report. https://lnkd.in/gxA-bYxz

34

There is a dangerous addiction in the Indian ISP space: The obsession with Coverage Maps. Running ISPs in India taught me a hard truth that few founders want to admit: We are often scaling our footprints faster than we can scale our reliability. We celebrate launching in 10 new cities. We high-five over "homes passed." We glorify the fiber mileage. But here is the reality check: If your churn rate is climbing as fast as your expansion rate, you aren't building a network. You’re building a leaky bucket. The real heroes of this industry aren’t the giants painting the map red. The heroes are the local ISPs who obsess over the boring, unsexy things: Standardizing the last-mile cabling. Having a technician at a customer's door in 30 minutes. Delivering 99.9% uptime, not just promising it. I’ve seen smaller players with zero marketing budget beat giants simply because they deliver Giant Trust.

45

3 Comments

There are a lot of different NaaS vendors out there ..but none are quite like Meter. The company builds its own hardware, software (mostly) and has an interesting take on SDN that it calls - Desired State Networking “Networking is going through actually a bigger change than people realize, because of many factors, including what’s possible in protocols, what’s possible in hardware and software, but also because fewer network engineers are actually coming into the industry compared to what’s needed,” Anil Varanasi, Meter co-founder and CEO, told me.

20

1 Comment

Final slide from my recent latency-related presentation - this one very much my own personal opinion with plenty of speculation. Fixed ISPs continue to deliver ever-better reliability, more bandwidth, better latency - at least to a demarcation at the edge of the home. But IMO user wants & needs have shifted and are now broader than the industry has fully recognized (though companies like CUJO AI® are exploring it). For example, a user may not be fully satisfied until they know that X application session on Y device and a point in time T is consistently great - across the full day, across all the devices they use, across all the apps they use, across the entire footprint of their home (or travels). Food for thought... #Bandwidth #Latency #NetworkResponsiveness #TemporalInternet #HolisticInternet #Broadband

73

5 Comments

𝐆𝐨𝐨𝐠𝐥𝐞 𝐭𝐨 𝐈𝐧𝐯𝐞𝐬𝐭 $𝟏𝟓 𝐁𝐧 𝐢𝐧 𝐀𝐈 𝐃𝐚𝐭𝐚 𝐂𝐞𝐧𝐭𝐫𝐞 𝐢𝐧 𝐕𝐢𝐳𝐚𝐠 Google Cloud CEO Thomas Kurian announced that the tech giant will invest $15 billion to build the world’s largest AI and data centre in Visakhapatnam, Andhra Pradesh. This hub will be Google’s biggest AI facility outside the US and will feature Tensor Processing Units (TPUs) for cutting-edge AI computing. “This gigawatt-scale AI hub represents a major step in supporting India’s digital infrastructure goals, and we’re grateful for the support of the Government of India and the state of Andhra Pradesh,” said Kurian. Google Cloud CEO Kurian made this announcement during Google’s event on “India’s AI-Powered Future and Advancing the Vision of Viksit Bharat 2047,” in the presence of Ashwini Vaishnaw, Minister of Electronics & IT; Nirmala Sitharaman, Finance Minister; and N. Chandrababu Naidu, Chief Minister of Andhra Pradesh. Google Cloud Thomas Kurian Ministry of Electronics and Information Technology Via: Outlook Business | Deepsekhar Choudhury (Story link in the comment)

21

1 Comment

Satya’s take on the Dwarkesh podcast on scaling AI infrastructure was refreshing. Their data center strategy is to “scale in time” vs “scale at once”. His take? Why build massive data centers at once and get locked in to one chip design that’ll be outdated in several years? “Microsoft data centers should be flexible enough to accommodate future generations of chips and workloads”. While at QUITE a different scale and stack, this is Casper Studios philosophy too. Instead of for data centers, we share the same insight for models, platforms and methods of implementation to clients. Hit singles and doubles. Scale when needed. You don’t need it all at once. We’re here to help with either when the time is right. I’m going to be guiding clients through a similiar mental model moving forward.

12