Normally, lowering risk means lowering expected return. Diversification is the exception: by holding many assets whose fortunes aren't perfectly correlated, you reduce the volatility of the whole portfolio without giving up its expected return, because individual disasters get averaged out. Harry Markowitz won a Nobel Prize for formalizing this, and it's the closest thing finance has to a free lunch.
The practical failure mode is concentration you don't recognize as concentration. Holding ten tech stocks isn't diversification — they crash together. Holding your employer's stock while employed there is worse: a single bad event takes your job and your savings simultaneously. Broad diversification means across companies, sectors, geographies, and asset classes. The uncomfortable corollary is that diversification guarantees you'll always own something that's performing badly — that's not a flaw, that's the mechanism working. If everything you own is going up together, you're not diversified.