For all intents and purposes, Vega was quite successful. It has found its way into products ranging from 35W £50 parts to 350W £500 parts, demonstrating an amazingly versatile balance of power & efficiency. It significantly outperformed the competition in its primary markets(Enterprise & servers, where pretty much all of AMDs recent profit has came from), shifted market share from NVidia to AMD on the desktop, and was consistently selling faster than it could be produced for most of its life. Heck, even Intel adopted Vega. It's important to remember a world exists outside the bubble of PC gaming, and it will always be a secondary market for any company capable of being successful in enterprise endeavours.
Even when companies do shift focus towards gaming, a company looking to carefully & reliably increase its profits would only really care about console gaming and mid/low end PC gaming, two markets AMD has fairly neatly wrapped up. High-end gaming GPUs are risky, low volume, low profit products, which is why both NVidia and AMD's high end cards are both now enterprise chips first with the gaming variants being cut-down second order products.
Turing & Vega both compete on very even footing in the market where they produce all their profit. Neither NVidia nor AMD seem to care about how they compete in the gaming market, hence why NVidia is flat-out refusing to release new products in AMDs target markets and vice versa. AMD doesn't have a reason to release new hardware to be more competitive, and neither does NVidia, as long as they both insist on playing it safe(But AMD isn't in a position to take risks and NVidia has never had a reason to).
