Silicon as software: why we invested in DualBird
Gil Dibner

We live in an era where the demand for data processing is exploding, driven by the voracious appetite of AI and analytics. Yet, the traditional levers for performance—optimizing software or simply throwing more capital at massive data centers—are hitting a wall.

As DualBird’s CEO Amir Gilad puts it, "Data processing is the biggest workload still stuck on general-purpose CPUs. It deserves purpose-built processors just like AI has GPUs."

DualBird represents one of the most compelling teams we’ve backed and one of the most ambitious technology bets we’ve ever made. Briefly put, the company believes that it can accelerate data operations by more than an order of magnitude by fusing semiconductor technologies with cloud-based, software-only deployment.

When we first met Amir Gilad and Gilad Tal, they were proposing something that sounded borderline impossible: delivering silicon-level performance improvements over the cloud in a "software only" way. In essence, they are building "silicon as a cloud-native software service."

Why we invested

Here are the core elements of our thesis on DualBird:

• Outstanding team. Amir Gilad (CEO) and Gilad Tal (CTO) are both exceptional founders who bring deep expertise in both hardware and software engineering. With co-founders Ehud Eliaz and Ohad Gamliel, they are semiconductor industry veterans with deep cloud software experience. It is exceedingly rare to find a team that truly understands the atomic level of silicon design and the operational complexities of modern cloud databases.

• Unique proprietary technology. What DualBird is doing is really, really hard. It is a multidisciplinary challenge that requires fusing two worlds that rarely speak the same language. It’s common knowledge that hardware-based acceleration can achieve 100x the performance lift of any software-based optimizations. The challenge has always been the logistical complexity of deploying that type of hardware efficiently. Achieving the impact of hardware with the deployment model of software has been a holy grail in the industry for years, but no one managed to achieve it. DualBird has found a way to leverage cloud-based FPGA technology (like Amazon F1/F2) to deploy software-based DB accelerators. Crucially, because these FPGAs are rented cloud resources, DualBird’s technology is by default co-located with customer workloads. This solves the massive logistical nightmare of traditional hardware acceleration. There is no shipping of boxes, no installation, and no complex integration. It operates as a plug-in for existing Apache Spark and Apache Iceberg environments, meaning deployment is seamless and latency is virtually non-existent.

• Significant performance and cost gains in a massive market. DualBird’s engine allows for 10-100x faster performance and 50-90% cost savings in a massive area of spending: data operations. These gains are orthogonal to software improvements—meaning they are robust and relevant even for the most sophisticated customers.The economic stakes here are enormous. McKinsey estimates that keeping pace with demand will require nearly $7 trillion in new data center investment by 2030. We cannot simply build our way out of this with more concrete and copper; we need efficiency at the silicon level. Imagine if all those computationally intensive data operations could be performed at a fraction of the cost.

• First ever. This has literally never been done. When the team at DualBird said they would deliver Hardware-style acceleration in a software-only way over the cloud, we had never heard that approach. We didn't know it was technically possible, and we certainly didn't have any idea what the ultimate business model would be. This is exactly the reason we got into venture—to back people building on the absolute cutting edge of what's possible.

How the deal came together

When we met the company, they were seeking to raise more than our fund could invest alone. Despite the high price of entry relative to our fund size, we conducted our diligence, built our conviction, and decided to move forward. We agreed with the founders to invest roughly half the round, and we committed to helping them secure the rest. Before any other investors had agreed to join, we signed final legals and wired our cash immediately. Ultimately, we were joined by our friends at Uncork Capital (Andy McLoughlin) and Bessemer Venture Partners.

Now, DualBird has finally announced total funding of $25M, with the most recent financing led by Lightspeed. It is a massive milestone for a radically innovative deep tech company, but for us, the excitement goes back to those early meetings where we realized we were looking at a team capable of breaking the wall between hardware and software.

Where we go from here

The first order of business for a company like DualBird is to prove that it can scale its business model. That challenge still lies ahead. (If you are running expensive data workloads—please contact the team or me directly! And if you want to be a major part of the GTM effort in a brand new category, let us know!)

But once that milestone is achieved, the road ahead is even more exciting. If DualBird can prove this model at scale, it could represent the opening play in an entirely new industry: the delivery of silicon-level efficiency via the cloud. Just as the "fabless" semiconductor model unlocked trillions of dollars of value over decades by separating design from manufacturing, it is possible that this "chipless" model will unlock trillions more by separating acceleration from physical deployment.

We are absolutely thrilled to be on this journey with Amir, Gilad, and the entire DualBird team as they rebuild the foundation of data analytics and AI infrastructure and figure out how to build an entirely new type of company. This is exactly what we mean by “hard tech, hard markets.”

The AI-Native Enterprise Playbook
Ten real-time observations on a rapidly evolving playing field

No More Painting by Numbers
It’s the end of the “SaaS playbook.

The Age of Artisanal Software May Finally be Over
Every wave of technological innovation has been catalyzed by the cost of something expensive trending to zero. Now that’s happening to software.

Founders as Experiment Designers
David on why founders should run everything as an experiment.

ENTERPRISE/TECH NEWS

Musical Chairs - Apple finally axed their embattled AI executive John Giannandrea. Although the Brit, who served as senior vice-president for machine learning and AI strategy is technically moving to an advisory role ahead of retiring in the Spring. The FT reports that “he will be replaced by former Microsoft executive Amar Subramanya, who leaves a job as a corporate vice-president for Microsoft six months after jumping ship from Google, where he worked on the Gemini chatbot.” They go on to say that Subramanya will report into the SVP of Engineering Craig Federighi, and that this could be another indication that Federighi is in contention for the CEO spot as the Board plans current CEO Tim Cook’s succession plan. 

An Open Field - The FT also reports that a recent report from MIT and Hugging Face found that China has now overtaken the US in ‘open’ AI model downloads. “Widespread adoption will confer outsized influence over AI’s future. China’s push to release open models comes in stark contrast to the “closed” approach of most of the biggest US tech companies such as OpenAI, Google and Anthropic.” Researcher Shane Longpre, MIT, commented on this as a paradigm shift, where Chinese companies were shipping models on a weekly or biweekly basis, with many different variations that users can choose from, rather than releasing a series of models every six months or year like US labs. “Other experts said that while China might be restrained in computing power, thanks to US export controls on powerful chips, the country has a wealth of homegrown researchers. This has allowed the country’s AI groups to be more creative in their approach to model development, using techniques such as distillation to create smaller yet powerful models. The country’s AI labs have also leaned in heavily into developing AI video-generation models.”

HOW TO STARTUP

GTMEng - A great listen from Lenny’s podcast with Vercel COO Jeanne deWitt (prev Stripe & Google) on ‘the future of AI-powered sales’. “You can’t really apply go-to-market engineering unless you have a point of view on what best practice should look like.” She advises founders to think about revenue operations earlier than you think, and that the same goes for implementing GTM Engineering as early as possible. 

It’s Real - Ilya Sutskever’s conversation on Dwarkesh Podcast - ‘We're moving from the age of scaling to the age of research.’ In the ‘hot mic’ conversation before they formally start the interview Sutsekever’s enthusiasm and amazement at the pace of the current moment is apparent and he comments “You know what’s crazy? That all of this is real.”

The “I Want” Song - Dave Kellog, now EIR at Balderton Capital, and previously GM of Service Cloud at Salesforce offers a lesson from the world of musical theatre, where in the first 10-20 minutes of the show the protagonist will articulate what they want in the form of song (after we’ve been introduced to the characters, but before we understand their motivations). He thinks this is a lesson for B2B software messaging: "Every buyer has an “I Want” song: usually unspoken, often half-formed, but always present. Yet most messaging fails to reveal it. Instead, we default to talking about our own technology, our architecture, our features, our AI, our category. We sing our song. When we should be singing theirs."

HOW TO VENTURE

Deal Sourcing - Early-stage VC sourcing is just different now....(and it's not because of AI). Charles Hudson’s reflections from his AGM are worth reading: "If there was an era when simply having an early signal put you at the front of the line for meeting with a new founder, that era is long over. In addition to early signal, your outreach needs to come with more than just a recognition that you know or suspect the person might be starting a new company - the best founders are discerning and being first or early doesn’t mean what it used to in a world where everyone has their own system for surfacing people who are on the verge of starting new companies."

Moonshot Space steps out of stealth after raising $12M led by Angular Ventures.

Gil Dibner talks about New Venture Swimlanes on The Random Walk’s latest podcast edition of Investors Who Talk Smart.

DualBird
DevOps Engineer (Tel Aviv / Haifa)

Firebolt
Software Engineer, Query Processing (Munich)

Motorica
Account Executive (Remote)