Maincode, the AI lab building Australia’s first sovereign LLM, is spinning up a venture capital arm its hopes will turn leaders in finance, healthcare and entertainment into disruptors.
Billionaire Stake founder Ed Craven has become the cornerstone investor of Maincode Ventures, a new outfit hoping to reimagine venture capital and accelerate the buildout of Australia’s sovereign AI capabilities.
The fund is an offshoot of Maincode, an AI lab funded by Craven that is building Australia’s answer to ChatGPT, and hopes to coax prospective entrepreneurs out of blue chip companies by reducing the downside risk of becoming a founder.
“Our thesis is that the future of AI is us [Australia] doing more ourselves,” said Maincode CEO David Lemphers. “If you want to disrupt healthcare, education, entertainment and finance, you need to go to the people working at [leading] organisations and give them everything they need to disrupt that organisation.”
Maincode Ventures’ proposal is unconventional. It takes 50 per cent of equity in all startups it invests in. In exchange, founders receive funding to build a team and operations, office space, free access to Maincode’s Matilda AI, and a salary.
That is far more equity than founders normally sell up front. Typically investors take around 20 per cent at the seed stage, with similar chunks of equity put on offer at Series A and B stages. It is also different to an accelerator like Y Combinator, where founders give up a portion of equity to go through a startup bootcamp and receive a small amount of funding.
Lemphers argues that this deal is fairer for founders. Maincode Ventures will cover much of their costs and provide the luxury of an “market” salary. Founders are free to sell parcels of their 50 per cent equity to new investors and, unlike universities or some accelerators, own all their IP.
“We talk to a lot of the VC funds in Australia and abroad, and at the end of the day all they can give you is a cheque,” Lemphers said. “Or you’re in a studio accelerator where you’re actually getting neither… you get a small cheque, they take a position on your cap table, they give you some network, but it’s really on you.”
Its model is similar to a venture studio, where founders hone their startup ideas and get access to internal engineering and product teams to get it off the ground. Venture studios have produced the likes of Snowflake (market cap US$116 billion) and Moderna (US$80 billion).
Craven is co-founder and key investor in Maincode, with his initial $30 million investment in the startup going towards securing a GPU cluster to train the Matilda large language model. Craven has issued a separate pool of funding for Maincode Ventures, though the company nor Craven would disclose how much.
Craven founded crypto casino Stake.com alongside Bijan Tehrani in 2016 and is now worth an estimated $2.2 billion.
Also unlike traditional VC firms, which fundraise for a period of time before closing the fund and investing from it, Maincode Ventures is an open structure that will continuously fundraise. The VC is also backed by other family offices and high net worth individuals, Lemphers said.
The current goal is to build 10-15 companies, with Maincode being the model itself. The startup, which employs around 15 people, has built a sovereign Australian AI model that runs on Australian infrastructure. The goal, Lemphers said, is to build similar capability across different industries.
“We’re having these conversations now with physicians, economists, people in entertainment, who are like, ‘I’ve got this insane idea, I need killer AI and I need to build a fairly significant sized team.’
“It might end up being a million dollars a month in payroll, but we’ll cover it. That’s the risk because we believe that’s where the future is going to be.”
Want to see more Forbes articles on your feed? Tap here to make Forbes Australia a preferred source on Google.
Look back on the week that was with hand-picked articles from Australia and around the world. Sign up to the Forbes Australia newsletter here or become a member here.
