HTTP/3 (Hypertext Transfer Protocol 3), will do away with TCP and instead use QUIC.
This follows from past when Google’s SPDY protocol was used as the foundation for HTTP 2.0 protocol.
What is QUIC?

QUIC or rather HTTP-over-QUIC is an experimental protocol developed by Google since 2013, which relies on UDP for data transmission instead of TCP. The key highlight of QUIC is because it is based on UDP, the number of round trips needed is reduced.
QUIC which stands for Quick UDP Internet Connections, improves upon SPDY (the HTTP 2.0 protocol) by reducing undesirable latency. This is achieved because a single lost packed in underling TCP connections stalls all of the streams over that connection. While a single lost packet for QUICK will only stall 1 connection our of several. With UDP, QUIC can support out-of-order delivery, so that a lost packet will typically impact (stall) at most one stream.
Why is TCP not good enough?
This is further exacerbated by the fact that TCP requires several greetings to establish a connection which requires several round trips increasing the latency.
For the simple reason that TCP focuses on reliability and perfect delivery of data over IP, that means, a single lost data packet might require a complete reset of the connection and a do over.
What next for HTTP/3?
While HTTP/3
Google wants QUIC to slowly replace TCP over time, and the adoption of slowly increasing with both Chrom and Opera supporting the HTTP rewrite HTTP-over-QUIC.
The IETF has provisionally approved to rename HTTP-over-QUIC as HTTP/3, however they are working on standardising QUIC, and presumably the standardised version will have deviations from the Google’s version, without affecting the core principles and benefits.