The market structure is pretty different. Restaurants have geographic barriers to entry - your restaurant is probably only serving customers within a ~20 mile radius. And the economics and business model are well-known: you know exactly how much rent is going to cost, how much labor is going to cost, how much food is going to cost, and how many tables you can turn over a night, and so you can build reasonable financial models for how much you might make.
Software is global, and is fundamentally an innovation business. Once you've written a piece of software that does something useful, you can sell additional copies at zero marginal cost. This tends to make software into a winner-take-all market: there is realistically only one Google, only one Facebook, only one Salesforce, one Amazon, etc. If you try to get into a known market, you are almost certain to fail, because you have to pay all the R&D costs that your competitor has already paid and they can just sell to the customers you would otherwise have gotten at close to zero cost. That means that successful software businesses are almost always doing something fundamentally new - either selling into a new market, or selling a new and different product into an existing market that has changed in some way. Banks are really bad at forecasting the success of new business models that have no financial data to go on - their whole core competency is evaluating financials, so if a company has no revenue but lots of expenses and an uncertain prospect of ever making money, it looks like a universally bad bet for a bank loan. The venture capital industry is all based around answering "How do we finance businesses where success is binary and information about whether the company will be successful is scarce?"